IBM says it has solved one of three key problems with making computer chip transistors from carbon rather than silicon. It’s a step closer to overcoming the current limitations on Moore’s Law.
The theory is that while silicon transistors have worked well for years, it’s getting harder and harder to make them any smaller and thus increase the overall capacity on a chip. The suggested alternative is a carbon nanotube, literally a rolled up sheet of carbon that is just one atom thick.
One of the key problems with using carbon nanotubes has been that the metal contacts needed to connect them in a transistor are relatively large. That makes it impossible to fit as many nanotubes on a chip as would otherwise be possible.
The solution turned out to be simple in concept, if intricate to turn into practice. Until now, nanotube transistors have worked by putting the contact along the tube’s length. IBM’s solution instead involved putting the contacts at the ends of the tube. That allowed the set-up to work with contacts that are 9 nometers long, compared with the equivalent contacts being 25 nanometers long on silicon chips.
With this solution, scalability is no longer a limiting factor in using carbon nanotubes. Two problems still remain however. One is that a nanotube can be formed in two ways, known as metallic and semiconducting. Only the semiconducting form is useful in a transistor, but filtering these out is still tricky.
The other problem is the sheer logistics of placing a tiny nanotube in the correct position on a chip, a task that needs to be repeated billions of times for each chip.