Carbon To Replace Copper On Semiconductor Superhighways
Copper has had plenty of time in the sun as a medium for electrons to travel to work each day. And although it has only been used in semiconductor interconnects for about 10 years, its days may be limited thanks to carbon. The debate on using carbon nanotubes (CNTs) for semiconductor interconnects has been ongoing for a while now, but recent research indicates that carbon offers first-class service to travel-weary electrons.
As semiconductors have gotten smaller, the copper wire used for interconnects has also shrunk, causing the resistance and heat to increase while degrading conduction. Consequently, electrons on busy highways are dealing with the equivalent of lazy construction workers that have decided to close off lanes, allowing fewer electrons through and leaving semiconductor vendors looking for alternatives to copper. Enter CNTs, which provide superior conductivity and mechanical integrity, says research from Rensselaer Polytechnic Institute (RPI).
RPI has been busy measuring and comparing key characteristics of both copper wire and CNT bundles as semiconductor interconnects. The modeling is quantum-mechanical in nature based on simulations run on RPI’s supercomputer—the most powerful university-based supercomputer in the world, according to the institute. The researchers also say that the resistance of CNT bundles is much less than that of copper nanowires, making them more ideally suited for interconnects, assuming the manufacturability Achilles’ heel of CNTs can be overcome.
“With this study, we have provided a roadmap for accurately comparing the performance of copper wire and carbon nanotube wire,” says Saroj Nayak, an associate professor in Rensselaer’s Department of Physics, Applied Physics, and Astronomy, who led the research team. “Given the data we collected, we believe that carbon nanotubes at 45 nm will outperform copper nanowire.”
The researchers chose quantum mechanical modeling instead of traditional empirical law-based modeling because of its accuracy. Quantum phenomena that don’t show up at the macroscale affect nanoscale interconnects. Empirical law-based modeling falls short when dealing with such atomic and subatomic phenomena, so it cannot predict the performance of copper nanowire. Quantum mechanic modeling is much more compute-intense, but allows for a more accurate model.
“If you go to the nanoscale, objects do not behave as they do at the macroscale,” Nayak said. “Looking forward to the future of computers, it is essential that we solve problems with quantum mechanics to obtain the most complete, reliable data possible.”
Yet many challenges lay ahead with CNTs (see “New Technologies Enable More Moore”, “Back To Nature For Next-Gen Semis”, and “So, What Was That Memory Technology Again?”. “More study will also be required,” Nayak notes, “to model and simulate the effects of imperfections in carbon nanotubes on the electrical resistance, contact resistance, capacitance, and other vital characteristics of a nanotube interconnect.”
Rensselaer Polytechnic Institute