IBM Scientist Unveils Carbon Nanotube Technology to Extend Moore’s Law

IBM Research scientist Qing Cao unveils a breakthrough method to interconnect carbon nanotubes.

The impending slowdown of silicon‑based chips has long prompted researchers to seek alternatives. IBM’s Qing Cao claims carbon nanotubes (CNTs) – cylindrical sheets of carbon only a few atoms wide – offer a path to smaller, faster, and more energy‑efficient transistors that can sustain Moore’s Law for decades.

Recognized as a Pioneer Under 35 by MIT Technology Review this year, Cao demonstrated how CNTs can be precisely aligned into dense arrays and then bonded to minute metal connectors. This technique, presented at TR’s EmTech in Cambridge, MA, paves the way for scaling CNTs to match today’s silicon chip dimensions.

What are silicon’s limits, and how do CNTs surpass them?

Qing Cao: Current silicon chips house billions of transistors at 22 nm, and we have already fabricated 7 nm nodes. However, silicon reaches a physical barrier around 5–6 nm due to quantum tunneling effects. CNTs, with diameters around 1 nm (four atoms across), enable devices at the 5 nm scale and beyond. At these dimensions, CNT transistors can operate twice as fast as silicon while consuming less than half the power.

Why haven’t CNT chips been commercialized yet?

QC: As device dimensions shrink, the interconnects that connect transistors must also become smaller. Reducing metal connectors below 10 nm dramatically increases resistance, undermining performance. Our team at the Thomas J. Watson Research Center has engineered a robust chemical bond between CNT ends and molybdenum wires, demonstrating that even 40‑atom‑wide metal connections preserve device performance.

Building a full CNT wafer remains a challenge. We have devised a self‑assembly process that arranges CNTs into side‑by‑side arrays across a wafer, enabling the integration of metal contacts with minimal resistive loss. This combination could yield chips that are smaller, faster, and more energy‑efficient than their silicon counterparts, effectively extending Moore’s Law.

When might we see these chips in consumer devices?

QC: We anticipate CNT‑based devices entering products within the next 10–15 years, sustaining Moore’s Law for at least the next two decades. Our long‑term vision is to pack 1 trillion transistors onto a single processor – more than the stars in the Milky Way.