Carbon nanotubes best for 3D electronics

Microelectronics International

ISSN: 1356-5362

Article publication date: 4 May 2012



(2012), "Carbon nanotubes best for 3D electronics", Microelectronics International, Vol. 29 No. 2.



Emerald Group Publishing Limited

Copyright © 2012, Emerald Group Publishing Limited

Carbon nanotubes best for 3D electronics

Article Type: Industry news From: Microelectronics International, Volume 29, Issue 2

Gothenburg, Sweden:

Researchers at Chalmers University of Technology have demonstrated that two stacked chips can be vertically interconnected with carbon nanotube vias through the chips. This new method improves possibilities for 3D integration of circuits, one of the most promising approaches for miniaturization and performance promotion of electronics.

3D integration is a hot field within electronics since it offers a new way to package components densely and thus build tiny, well-functioning units. When stacking chips vertically, the most effective way to interconnect them is with electrical interconnects that go through the chip (instead of being wired together at the edges) – what are known as through-silicon vias (TSVs).

The industry thus far has primarily used copper for this purpose; however, copper has several disadvantages that can limit the reliability of 3D electronics. Another major issue involves cooling when the chips get hot. The excellent thermal qualities of carbon nanotubes can play a decisive role in this respect.

Thus, a research team at Chalmers is working with carbon nanotubes as conductive material for TSVs. Carbon nanotubes – or tubes made of graphene whose walls are only one atom thick – are going to be the most reliable of all conductive materials if it is possible to use them on a large-scale. This is the opinion of Kjell Jeppsson, a member of the research team.

“Potentially, carbon nanotubes have much better properties than copper, both in terms of thermal and electrical conductivity”, he says. “Carbon nanotubes are also better suited for use with silicon from a purely mechanical point of view. They expand about the same amount as the surrounding silicon while copper expands more, which results in mechanical tension that can cause the components to break.”

 Figure 1 Two chips have interconnects that are filled with thousands of
carbon nanotubes

Figure 1 Two chips have interconnects that are filled with thousands of carbon nanotubes

The researchers have demonstrated that two chips can be vertically interconnected with carbon nanotubes by TSV interconnects, and that the chips can be bonded. They have also demonstrated that the same method can be used for electrical interconnection between the chip and the package (Figure 1).

PhD student Teng Wang – who defends his thesis on December 12 – has worked on production. He has developed a technique to fill TSVs with thousands of carbon nanotubes. The chips are then bonded with an adhesive so that the carbon nanotubes are directly contacted and can thus conduct current through the chips.

“One difficulty involves producing carbon nanotubes with perfect properties and with the length we need to go through the chip,” he says. “We have produced tubes that are 200 μm long, which can be compared to the diameter which is only 10 nm. Their properties, however, are not yet perfect.”

For the method to be transferred to industrial production, manufacturing temperature needs to be reduced to a maximum of 450°. This is a great challenge since carbon nanotubes are currently “grown” at a minimum of 700°.

If successful, entirely new possibilities will arise for future shrinking of electronics – not least in terms of improved performance. The 3D integration using TSVs provides significantly quicker signal transfers than traditional integration where chips are placed next to each other. Furthermore, TSVs with carbon nanotubes provide less expensive production compared to the current technology that uses copper interconnects.

“There are several projects involving 3D integration underway in the industry, but there are potential problems with both cooling and reliability since they use copper,” says Kjell Jeppsson. “If our method works on a large-scale, I believe it will be in production within five years.”

Related articles