Copper interconnects carry current in today's integrated circuits, but in the nanometer-size future, the metal just won't do the job. At this month's IEEE 2010 International Electron Devices Meeting, in San Francisco, European researchers plan to announce that they're one step closer to a replacement. Nanotubes made of carbon, if grown in dense bundles, can transport large quantities of charge through tiny channels reliably. As part of the European ViaCarbon project, a team led by Jean Dijon, the head of nanotube research at the French government research organization CEA LITEN in Grenoble, says they've grown the densest bundles yet, packing 2.5 trillion carbon nanotubes per square centimeter. The density of their interconnects is within an order of magnitude of what's needed for replacing copper. In the future, such bundles have the potential to exceed copper's current-carrying capabilities by a factor of 100.
Copper needs a replacement particularly in the narrow pegs, called vias, that connect the silicon surface to the chips' wiring and connect one layer of wiring to another. According to the 2009 International Technology Roadmap for Semiconductors, engineers predict that as the features on chips shrink, not only will copper vias be more difficult to manufacture and suffer from more resistance, but by 2015 they may not work at all.
Vias start as holes etched into a layer of dielectric. Depositing copper ions from an electrolyte fills the holes, but at the nanometer scale, the metal could cling to the walls, leaving gaps. Also, the vias' walls must be lined with a metal nitride barrier to keep copper from seeping into the surrounding dielectric. The liner wastes space and increases resistance.
Engineers are also concerned that copper vias will prove unreliable at future nanometer-scale dimensions. Skinnier copper means that current density and resistivity both increase. Together, those factors cause the interconnects to heat up and break. "For each new generation, as dimensions decrease, current densities increase and will soon approach the copper limit," says Murielle Fayolle-Lecocq, a microelectronics engineer who worked on the new nanotube vias at France's CEA-LETI.
Carbon nanotubes have the perfect shape for fitting into future vias' voids: "Naturally, they're tall and thin," says the leader of the ViaCarbon project, John Robertson, who is an electrical engineering professor at the University of Cambridge, in England.
The team's goal was to grow dense bundles of narrow nanotubes. Although nanotubes are usually championed for their ability to carry charge long distances with little resistance, quantum interactions in individual nanotubes can actually give them more resistance than the copper interconnects they're meant to replace.
Bundles of parallel nanotubes can solve that problem by providing multiple channels for electricity to flow. That lowers the nanotubes' effective resistance, making them seem like many individual resistors connected in parallel, says Robertson. Using nanotubes with only a few layers and carefully selecting the material on which they grow will ensure that the tubes have a small diameter and thus increase the possible bundle density.
If packed densely enough, these bundles can carry about 1 billion amperes per square centimeter, around 100 times copper's limit. But they haven't reached that ideal yet. Although the team has packed the tubes 10 times as densely as others, Robertson says, to match copper's conductivity they will need to increase the density by another factor of 10, to around 30 trillion nanotubes per square centimeter.
Adrian Ionescu, head of the Nanolab at École Polytechnique Fédérale de Lausanne and a contributor to the nanotube via project, adds that the team will need to improve the tubes' contact with the metal horizontal interconnects in integrated circuits. Ionescu believes that in the future, nanotubes—or a combination of nanotubes and graphene sheets—could replace the copper in even these horizontal interconnects.
"They still have to increase the density by quite a bit," says Kaustav Banerjee, director of the Nanoelectronics Research Lab at the University of California, Santa Barbara, who was not involved in the research. But, he adds, "this is definitely a significant improvement and very positive news."