GA Tech: Graphene could replace Cu for IC interconnects


Researchers at Georgia Tech say they have experimentally demonstrated the potential for graphene to replace copper for on-chip interconnects and help extend performance scaling for silicon-based ICs.

Their work, published in the June 2009 issue of the IEEE journal Electron Device Letters, shows resistivity in graphene nanoribbon (GNR) interconnects as narrow as 18nm, “comparable to even the most optimistic projections for copper interconnects at that scale,” and “probably already out-perform copper at this size scale,” according to Raghunath Murali, a research engineer in Georgia Tech’s Microelectronics Research Center, in a statement.

A key problem with copper interconnects is that at nanoscale-dimensions conductance is affected by scattering at the grain boundaries and sidewalls, Murali explained. “These add up to increased resistivity, which nearly doubles as the interconnect sizes shrink to 30nm.” At the ~20nm scale the increased resistance would offset performance increases and negates gains made in higher density???a roadblock to performance increases, if not to actual scaling, he noted. “We would be giving up so much in terms of resistivity, we wouldn’t get a performance advantage,” he said. And thus, the search for evaluating different materials for interconnects.

SEM image showing 22nm wide graphene nanoribbons between the middle electrode pair. (Source: Georgia Tech/Raghunath Murali)
Click here to enlarge image

In their work (Murali, Kevin Brenner, Yinxiao Yang, Thomas Beck, and James Meindl), flakes of multilayered graphene were removed from a block and placed in an oxidized silicon substrate; on that they constructed four electrode contacts using e-beam lithography, and then devices with parallel 18-52nm wide nanoribbons. 3D resistivity of the nanoribbons on 18 different devices was measured using standard analytical techniques at room temperature. Results indicated the best GNRs possessed conductivity equal to that predicted for Cu interconnects at the same size. Average resistivity at a given linewidth was found to be about 3?? that of a Cu wire; the best GNR had a resistivity comparable to Cu.

From the paper abstract:

    The conductivity is found to be limited by impurity scattering as well as line-edge roughness scattering; as a result, the best reported GNR resistivity is three times the limit imposed by substrate phonon scattering. This letter reveals that even moderate-quality graphene nanowires have the potential to outperform Cu for use as on-chip interconnects.

Moreover, they say, the comparison was between “non-optimized” graphene and “optimistic estimates” for Cu, and without using “especially clean processes.” A key property of graphene is its ballistic transport (low resistance), but the researchers noted that actual conductance is limited by scattering from impurities, line-edge roughness, and substrate phonons (vibrations in the substrate lattice). “With our straightforward processing, we are getting graphene interconnects that are essentially comparable to copper,” noted Murali. “If we do this more optimally, the performance should surpass copper.”

The research was supported by the Interconnect Focus Center (one of the Semiconductor Research Corp./DARPA Focus Centers) and the SRC’s Nanoelectronics Research Initiative through its Institute for Nanoelectronics Discovery and Exploration (INDEX). ??? J.M.