The key to Moore’s law is the ability to incorporate ever-smaller feature sizes into each new chip generation. While the exponential progress ensconced in the “law” has slowed down in the last decade, the payoff is still compelling enough that chip engineers will go to great lengths to forestall its long-predicted demise. One of these workarounds is to expand transistors into the 3rd dimension, which has the effect of speeding switching while reducing power. On the manufacturing side, a technique called self-assembly holds promise as a way to achieve smaller circuit elements by getting molecules to automatically arrange themselves into tiny but useful patterns.
According to a recent article at MIT Technology Review by guest contributor Katherine Bourzac, researchers at IBM have combined these approaches to create the first 3-D transistor made with molecular self-assembly.
As Bourzac explains, one of the primary enablers for chipmaking, photolithography, has hit a roadblock when it comes to the fastest microchips. Generally regarded as acceptable down to 14-nm, conventional photolithography is expected to become too expensive and complex past that point due to limits imposed by the wavelength of light.
The technique that IBM researchers employed involves solutions of compounds known as block copolymers that are coaxed to assemble themselves into complex structures. In this way, it is possible to create patterns that are much denser than what would be achievable with lithography. The technique is directed at the smallest elements of the integrated circuit that are most difficult to achieve with conventional methods, i.e., the channels in silicon transistors and the fins in 3-D transistors. For the rest of the circuit, standard manufacturing technologies would still suffice.
The IBM research group used photolithography to create deep, parallel trenches in a silicon wafer. The trenches guide the assembly of block copolymers, which then act as a template for a chemical process that etches even tinier features. The end result was a working device with transistor fins smaller and more densely-packed than would be possible with just lithography. The features were just 29 nanometers apart, much smaller than the 80 nanometers that is currently possible, writes Bourzac.
There is a lot of buzz around self-assembling circuits in the chip industry, where it is seen as a potential Moore’s law extender. However, according to Kowk Ng, director of nanomanufacturing at the Semiconductor Research Corp. (interviewed for the source article), the process is still open to defects that will need to be controlled before it is suitable for large-volume production.