Carbon Nanotube 16-bit Microprocessor Takes Computing Beyond Silicon26/08/2019 “We always hoped that something like this could be built – now we know that it can be built,” says Max Shulaker, professor at MIT and corresponding author on this latest report. Carbon nanotubes have been touted as a potential successor to silicon technology that could improve energy efficiencies by an order of magnitude. Now Shulaker and his team in Department of Electrical Engineering and Computer Science, alongside researchers at Analog Devices, Inc.(ADI) also in Massachusetts USA, have taken on a series of challenges that have hampered carbon nanotube (CNT) computers since the first carbon nanotube transistors were reported in the late 1990s.
The result is a 16-bit microprocessor comprising more than 14,000 complementary metal–oxide–semiconductor CNFETs that runs standard 32-bit instructions on 16-bit data and addresses, putting CNT microprocessor technology firmly on the map. The device is based on the RISC-V instruction set processor, which is open-source and commercially available, and the design and fabrication uses industry-standard design flows and processes.
What’s wrong with silicon?
Ever since Robert Noyce produced the first integrated circuit (IC) on a silicon chip in 1959, research and development has pushed processing powers and efficiencies higher and higher. Gordon Moore, Intel co-founder alongside Noyce, identified the trend as the number of transistors per square inch on an IC doubling every two years. However broaching nanoscale feature sizes has heralded a number of issues associated with leakage currents and the electronic property limitations of materials for gating and channelling. The apparent limit to “Moore’s Law” with silicon technology has prompted researchers to hunt for alternatives.
Carbon nanotubes – rolled sheets of honeycomb hexagonal carbon lattices – have attracted a lot of interest in this endeavour on account of their inherent nanoscale dimensions and impressive electronic properties, which can be semiconducting or metallic depending on the axis along the lattice that the sheet rolls up (the chirality).
What’s wrong with carbon nanotubes?
It was not long after the discovery of carbon nanotubes that people began to recognize their potential as “molecular” wires. However, their attractive attributes come with a number of caveats. They are prone to aggregating into bundles that kill the transistor performance, synthesizing nanotubes with specific chiralities remains impractical for IC purposes, and controlling the transistor type to produce the transistors with the complementary n- and p-type polarities central to CMOS technology is similarly problematic. The researchers identified a series of solutions to these issues: RINSE (removal of incubated nanotubes through selective exfoliation), MIXED (metal interface engineering crossed with electrostatic doping) and DREAM (designing resiliency against metallic CNTs).
RINSE tackles the aggregation issue, where previous efforts have either left some bundles or removed dispersed nanotubes (or both). Key to the success of Shulaker and colleagues is to spin coat an adhesive on the wafer so that sections of CNT do not come away along with the aggregated bundles during the subsequent sonication.
Christian Lau, a PhD student in Shulaker’s group who came up with the RINSE procedure, also came up with MIXED, which involves engineering the stoichiometry of the oxide used to encapsulate the CNT field effect transistors, as this controls the doping of the CNTs. They also engineer the contacts to optimize the p- and n-type transistors by using metals with complementary work functions (that is, higher for p-type and lower for n-type).
The final conundrum to resolve was the problem of metallic CNTs, which lead to leakage currents and ultimately interfere with the logic behaviour of the gates and undermine the performance. In fact the tolerance to metallic CNTs is so low that the circuits require semiconductor CNTs with a purity of 99.999999%. However what MIT post doc Gage Hills figured out was that certain logic gate combinations were more susceptible to the accidental metallic CNTs than others. He found that by designing the circuits to avoid certain pairings they could outsmart the pernicious effects of rogue metallic CNTs.
The researchers programmed the microprocessor to print the words: Hello, world! I am RV16XNano, made from CNTs. “It was certainly exciting when it worked – it was something all of the many of co-authors and I spent many years working towards,” says Shulaker. Hills adds, “My favourite part of being a part of this project is that it brings together researchers with expertise across a wide range of backgrounds, including material synthesis, physical fabrication, circuit design, computer architecture, and applications. Without innovations and support from the team in all of these areas together, it wouldn’t have been possible to demonstrate the modern microprocessor built entirely out of carbon nanotube field-effect transistors.”
At 1.5 micrometres, the 16-bit microprocessor is comparable with Intel’s silicon-based 80386 processor released in 1985. Having demonstrated that CNTs can produce sophisticated working microprocessors the researchers are now focusing on realizing the promised performance gains.
Of course, what works in a lab doesn’t always make its way onto the high street. By working closely with several commercial partners – collaborators from ADI co-authored the report of the work – the team have a good view of industry requirements to steer the research in the right direction for the transition from “academic lab to a commercial fab”.
“ADI’s success has always been built upon pursuing innovation at all levels of our organization and working with our larger ecosystem to continuously redefine what is possible,” says Vincent Roche, CEO of ADI. “Working with Prof. Shulaker’s group at MIT on carbon nanotubes is the perfect embodiment of this philosophy, and we are excited by its tremendous promise.”
Source: https://bit.ly/2m5z1rn, via Physics World