The current technique for mass production of computer chips is photolithography where light is used to etch the circuits. A material called a resist is placed on the layers of the chip, one at a time, and is exposed to a specific pattern of light, causing it to react. The layer is then cleaned, removing all of the resist and the material beneath the unexposed resist. The exposed resist protects the material it is on top of from the cleaning process. Repeating this several times results in the 3D structures within a computer chip. While this method has served us well for fifty years it has one flaw; it is reliant on the wavelength of light used. As the wavelength of the light gets smaller and smaller, the optics get more complicated, and the light source becomes less efficient. The smallest chip features photolithography has been able to resolve are 25nm wide. A team at MIT will soon have an article published on a way to produce features as small as 9nm at high speed.
The MIT method uses a modified form of electron beam (e-beam) lithography which operates in much the same way as photolithography. E-beam lithography is already used to prototype computer chips, but has some drawbacks to prevent it from being used in mass production. One issue is that a single beam will take a long time to etch all the circuitry, especially compared to photolithography that can expose all the resist on a chip at the same time. To solve that problem, a system using multiple beams could be used, such as those built by Mapper, a Netherlands company. One such machine built by Mapper has 110 parallel e-beams.
Another issue is the time it takes an electron beam to expose the resist, and this is what MIT directly addressed. The team used a thinner layer of resist and added table salt to the resist, to make it react better. This is very important because the easier it is for the beam to develop the resist then fewer electrons are needed, and so is less shielding in the device.
This technology isn’t ready to make our off-the-shelf electronics, yet, but then we aren't to 25nm yet either.