This is a piece I wrote for Materials Today. It originally appeared in the November 2013 issue but can be read here for free 🙂
A team from Stanford have built the world’s first functioning computer based on carbon nanotube (CNT) transistors.
The need for smaller electronic devices has driven the semiconductor industry towards the miniaturization of components, and has brought about reduced costs and major improvements in computational power and energy efficiency. But for silicon-based components, smaller, cheaper, faster also means hotter – as components shrink, and more transistors packed on each chip, the power density increases, and they generate more heat.
Digital circuits based on semiconducting carbon nanotube (CNT) transistors have the potential to outperform silicon by improving the energy–delay product, a metric of energy efficiency used in logic systems, by more than an order of magnitude. It is almost 15 years since carbon nanotubes were first used to produce transistors, where CNTs replaced silicon as the channel material within a MOSFET. But in a major development published in Nature, a team from Stanford has built a fully-functioning computer, built entirely from these CNT-based transistors.
With Cedric, as the system has been dubbed, Stanford have overcome the inherent difficulties in using CNTs that have limited their practical use as transistors. In their two-pronged attack, the researchers removed any metallic CNTs and accounted for any misalignment of the tubes. By biasing the gate, and pulsing a large current through the system, the semiconducting CNTs are retained while Joule heating causes the metallic CNTs to vaporise. The team also developed an algorithm and fabrication technique to identify and isolate those misaligned CNTs that would cause logic fail.
The CNT computer is capable of multitasking – running a basic operating system that can switch between counting and number sorting. And their system can also run MIPS – a commercial instruction set developed in the 1980s. Cedric is a Turing Complete, meaning that it could be used to solve any computational problem. It consists of 178 CNT transistors, meaning that it operates on one bit of data, and can count to 32. But this limit is not a physical one; it is due to the scale of the university chip fabrication facilities. Their design is compatible with current industry processes, so can be scaled up.
This paper demonstrates the promise of CNTs for use in complex computing systems. The team are confident that although large-scale development may be years away, Cedric offers a major breakthrough in low-power nanoscale computing, and may herald the end of silicon.
Nature (2013) doi:10.1038/nature12502