The obvious way to continue shrinking chip features would be to use beams of electrons to transfer mask patterns to layers of photoresist. But unlike light, which can shine through a mask and expose an entire chip at once, an electron beam has to move back and forth across the surface of a chip in parallel lines, like a harvester working along rows of wheat. “It’s like the difference between writing by hand and printing a page all at once,” says Karl Berggren, the Emanuel E. Landsman Associate Professor of Electrical Engineering, who along with Caroline Ross, the Toyota Professor of Materials Science and Engineering, led the new work. The slow, precise scanning of electron-beam lithography makes it significantly more expensive than conventional optical lithography.
Berggren and Ross’ approach is to use electron-beam lithography sparingly, to create patterns of tiny posts on a silicon chip. They then deposit specially designed polymers — molecules in which smaller, repeating molecular units are linked into long chains — on the chip. The polymers spontaneously hitch up to the posts and arrange themselves into useful patterns.
The trick is that the polymers are “copolymers,” meaning they’re made of two different types of polymer. Berggren compares a copolymer molecule to the characters played by Robert De Niro and Charles Grodin in the movie Midnight Run, a bounty hunter and a white-collar criminal who are handcuffed together but can’t stand each other. Ross prefers a homelier analogy: “You can think of it like a piece of spaghetti joined to a piece of tagliatelle,” she says. “These two chains don’t like to mix. So given the choice, all the spaghetti ends would go here, and all the tagliatelle ends would go there, but they can’t, because they’re joined together.” In their attempts to segregate themselves, the different types of polymer chain arrange themselves into predictable patterns. By varying the length of the chains, the proportions of the two polymers, and the shape and location of the silicon hitching posts, Ross, Berggren, and their colleagues were able to produce a wide range of patterns useful in circuit design
One of the two polymers that the MIT researchers used burns away when exposed to a plasma (an electrically charged gas), while the other, which contains silicon, turns to glass. The glass layer could serve the same purpose that a photoresist does in ordinary lithography, protecting the material beneath it while that around it is etched away.
Since Berggren and Ross’s technique requires no such channels to guide the self-assembling molecules, it reduces the need for electron-beam lithography. According to Herr, “That will save tremendously in terms of throughput” — that is, the efficiency with which chips can be manufactured.
Much more research is required, however, before self-assembling molecules can provide a viable means for manufacturing individual chips. Nearer term, Berggren and Ross see the technique’s being used to produce stamps that could impart nanoscale magnetic patterns to the surfaces of hard disks, or even to produce the masks used in conventional lithography: today, state-of-the art masks for a single chip require electron-beam lithography and can cost millions of dollars. In the meantime, Ross and Berggren are working to find arrangements of their nanoscale posts that will produce functioning circuits in prototype chips, and they’re trying to refine their technique to produce even smaller chip features.
Templated self-assembly of block copolymer thin films can generate periodic arrays of microdomains within a sparse template, or complex patterns using 1:1 templates. However, arbitrary pattern generation directed by sparse templates remains elusive. Here, we show that an array of carefully spaced and shaped posts, prepared by electron-beam patterning of an inorganic resist, can be used to template complex patterns in a cylindrical-morphology block copolymer. We use two distinct methods: making the post spacing commensurate with the equilibrium periodicity of the polymer, which controls the orientation of the linear features, and making local changes to the shape or distribution of the posts, which direct the formation of bends, junctions and other aperiodic features in specific locations. The first of these methods permits linear patterns to be directed by a sparse template that occupies only a few percent of the area of the final self-assembled pattern, while the second method can be used to selectively and locally template complex linear patterns.