Increasingly, quantum computers are predicted to be the next great leap in computational power — but in reality they are more likely to be the next next great leap. Right now we have to tailor experimental quantum chips to their particular mathematical process of interest, literally build them to solve a specific problem; today’s silicon solutions will reach the peak of their potential long before we can go buy Intel or AMD’s new plug-and-play quantum processor.
We need something that can continue to provide increasing computational power in the time between silicon’s peak and quantum’s grand beginning. One possible solution is graphene, a material that could dramatically increase computer performance without overturning the foundations of computer engineering.
Far from the conceptual quagmire of the quantum world, the mechanics of graphene processors are quite familiar to the silicon generation. Experimental graphene devices have already been clocked at hundreds of gigahertz, and workarounds exist for the material’s inconvenient lack of a bandgap for on-and-off switching. The main problem with graphene processors at this point is actually making them, working with the incredibly small-scale features that give graphene its miracle properties. It’s not easy to make a ribbon just a few dozen atoms wide, a single atom thick, and with a perfectly uniform honeycomb structure. Despite several years of intense research, there are still only tentative solutions on the table.
This week, Stanford University scientists detailed a new way to produce graphene ribbons, and from them working graphene transistors. The technique used DNA to provide a scaffold for the graphene synthesis, and some clever chemical tricks to provide the carbon atoms that make up the final product.
DNA’s simple four-base binding and organizational system lets scientists quickly and accurately create their basic graphene template. They dipped a platter of silicon into a solution rich in their engineered DNA strands, then “combed” the strands with molecular tools to stretch them out straight and uniform. The pattern became a convenient physical precursor to graphene, and treatment with a copper salt solution turned the DNA into a helpful chemical precursor as well.
Graphene is one of the purest carbon substances possible, up there with diamond in terms of homogeneity. When heated and exposed to the simple hydrocarbon methane, the copper-treated DNA double helix can donate some of its carbon atoms to create that pure-carbon honeycomb structure. This new technique is still in its infancy, creating flawed ribbons that can only be called “graphitic” due to small bunched regions that defy graphene’s uniform perfection.
Still, the products were pure enough to support the creation of actual graphene transistors. Graphene transistors have the potential to be scaled down much further than their silicon-based competitors, and can signal much more quickly, while consuming less power. These are the three basic requirements for improving chip technology: get more, faster transistors into an existing space while using no additional energy. Reliably (and cheaply) creating a high-quality chip of tightly packed graphene units is a major goal in computer engineering today.
The researchers note that this technique is easily scalable and ready to be adapted to the large-scale manufacturing techniques that could see a theoretical graphene processor in the consumer price range. Aside from the actual formation of graphene on this scaffold, all the lab techniques used here are well worn; dipping platters and taming wild strands of DNA are old hat for many areas of science, though the manufacturing industry could take some time to adjust to DNA’s fragile nature.
Unless we want to utterly stop with the historically unprecedented rate of progress we’ve enjoyed for the last fifty years or so, we’ll need a more proximate solution than the quantum computer. A graphene microchip could provide this, increasing performance without overturning any paradigms, supporting our increasing computational demands while still working in the familiar language of ones and zeroes.