Extending Moore’s Law
ALL of us have been beneficiaries of Moore’s Law, which dictates that every 18 months or so, the number of transistors that can be placed onto silicon chips, on an integrated circuit, would double. This is the reason why computers get faster and more powerful even though computer prices stay steady over the years.
This was a trend noticed by Intel co-founder Gordon Moore back in 1965. It’s amazing that Moore’s Law has held up as long as it has. But this won’t last forever and in fact, there has already been a slowdown due to the physical limitations of silicon.
Carbon nanotubes (CNTs), which I wrote about last week as the potential material to be used in space elevators, may be the saviour of Moore’s Law as well. Not only are CNT transistors faster than silicon ones, they’re believed to be as many as 10 times more energyefficient. This is important for the battery life of laptops, tablets and mobile phones.
As its name suggests, CNTS are tiny tubes of atomlevel, thin carbon sheet, and they happen to be excellent semiconductors, which means they can both conduct electricity as well as shut it off (just like silicon).
CNTs are a relatively new phenomenon. IBM made the first CNT transistors about 20 years ago but so far, attempts to make processors with CNT on a large scale have faced great challenges.
One key problem is that when CNTs are made, two types are created. The first are semiconductors which are ideas for making integrated chips. But the second are metallic in nature and conducts electricity like a wire (making it unsuitable
for integrated chips). Another issue is that to make the chips, a single layer of CNTs need to be deposited on top of a wafer but this is difficult to do because CNTs have a natural tendency to clump together.
A team at MIT, headed by Max Shulaker, has managed to overcome these problems and even managed to build a prototype 16-bit microprocessor with over 14,000 CNT transistors. This basic processor successfully executed a simple programme that produced the message: “Hello, World! I am RV16XNano, made from CNTs.”
The first problem — intermixing of semiconductor and metallic CNTs — was due to purity issues. The researchers calculated it would take a purity of 99.999999 per cent to avoid this intermixing. However, their current process is able to yield only purities of 99.99 per cent.
Instead of trying to attain that unattainable level of purity, they relied on the fact that different combinations of logic gates — i.e. groups of transistors designed to carry out a specific operation — were more susceptible to metallic CNTs than others. So, they create a set of design rules that avoid vulnerable combinations, allowing them to still build functional circuits at only 99.99 per cent purity. They called this process “DREAM” (Designing Resiliency Against Metallic CNTs).
“The ‘DREAM’ pun is very much intended because it’s the dream solution,” says Shulaker.
As for the clumping problem, they coated a 150mm wafer in a polymer and then employed an exfoliation process that washed away the clumps in stages. This stripped off clumps, leaving behind the single layer of CNTs needed to make the chip work. They called this process: “RINSE” (Removal of Incubated Nanotubes through Selective Exfoliation).
CHALLENGES AND SOLUTIONS
There was also one other challenge the researchers had to tackle. Computing requires two types of transistors: “N” types, which turn on with a 1 bit and off with a 0 bit; and “P” types, which do the opposite. Traditionally, making the two types out of CNT has proven to be challenging, often yielding transistors that vary in performance.
To resolve this problem, they attached metals to each transistor (platinum or titanium) which allows them to fix that transistor as either P or N. Then, they coated the transistor in an oxide compound, which allows them to tune the transistors’ characteristics for function and optimisation. They called this MIXED (Metal Interface Engineering Crossed with Electrostatic Doping”).
It should be said that these are still early days. The 16-bit microprocessor made with CNTs is a great achievement but in terms of performance, it’s comparable to Intel’s 80386 processor released in 1985. It’s really just a proof of concept, to show that it can be done, but it’s still quite an achievement.
“This is by far the most advanced chip made from any emerging nanotechnology that is promising for high-performance and energy-efficient computing,” said MIT’s Shulaker. “There are limits to silicon. If we want to continue to have gains in computing, carbon nanotubes represent one of the most promising ways to overcome those limits.”
For them to be able to build a chip that goes from “Hello, World”, to one that can do the kind of things today’s computers can handle, it will require processors with billions of transistors on them. “We think it’s no longer a question of if,” Shulaker said of commercially available CNT processors, “... but when.” He believes this could happen within five years.