Although there is a lot of life left in silicon devices, the day will come when silicon transistors cannot be miniaturized any further. Using methods from chemistry and electrical engineering, Jeff Bokor and Ali Javey, a chemist by training, are exploring new materials that may take silicon's place. "As transistors approach the molecular scale, chemistry is becoming important for controlling their properties," says Javey, who arrived at Berkeley in fall of 2006.
EECS Professor Ali Javey, with a cryogenic probe station, which is used to measure the electrical properties of devices. (Photo by Peg Skorpinski) The most promising silicon replacement is carbon nanotubes—long tubes of pure carbon that are only a few atoms wide and resemble a rolled-up sheet of graphite, yet are as stiff as diamond. Because of their unusual electrical properties, they "have emerged as one of the most remarkable materials for nanoelectronics,” says Bokor, who directs the Nanofabrication Facility at the Molecular Foundry, a nanoscience research center at Lawrence Berkeley National Laboratory. “You have what is in essence a single molecule of pure carbon, a rolled-up sheet with a diameter in the one-nanometer ballpark," he adds. "That’s a very interesting size for lots of applications.”
In ordinary crystalline materials, electrons bounce around a lot as they encounter impurities or defects in the lattice or simply bump into atoms that are vibrating thermally. While a graduate student at Stanford, Javey demonstrated that electrons in a nanotube move in a “ballistic” trajectory, meaning that they lose almost no energy as they travel. This translates to a higher current-carrying capacity and therefore faster switching speeds. Moreover, because there is no energy loss in the channel, nanotube devices can operate at lower voltages, potentially enabling high-speed and low-power electronics.
At Stanford, Javey succeeded in building cutting-edge carbon nanotube transistors with 10 to 30 times the current-carrying capacity of silicon. Still, there are many obstacles to overcome before carbon nanotubes can replace silicon. The diameter of a nanotube is difficult to control, and nanotubes with slightly different diameters have very different electrical properties. Also, nanotube transistors can be made only in relatively small numbers, and it is very difficult to place them accurately. “You need to position them within a few nanometers,” says Javey. “We’re really far away from that. We can get within hundreds of nanometers at best.”
Until these issues are addressed, Bokor believes that carbon and silicon will work together on hybrid chips, with the silicon transistors wired together in a conventional circuit, keeping tabs on the electrical properties of the carbon ones. Initially this would be necessary because those properties are not controlled well enough by the manufacturing process, but later it could be turned to advantage, enabling the nanotubes to become sophisticated chemical sensors. Bokor's lab has already manufactured chips with up to 4000 silicon transistors and 2000 carbon nanotube transistors.
Javey, meanwhile, is bridging chemistry and electrical engineering to address various issues in shrinking circuitry. For instance, traditional fabrication methods start with a block of material and drill it down. Javey is looking at ways to synthesize nanotubes and other nanomaterials for electronics from the bottom up. "Is it more logical to synthesize for the bottom up rather than trimming from the top down?" he asks. "That is a direction we are exploring." Synthesizing electronics would have the added advantage of enabling new applications, Javey notes, such as carbon-based devices where the various electronic components are printed in hundreds or thousands of layers that can communicate directly with one another.
Javey also plans to construct nanowires—nanotubes filled with silicon or gallium nitride—using bottom-up synthesis. “A carbon nanotube is a better transporter of charge, but nanowires would be larger and therefore easier to control for placement and assembly,” Javey says. “But no one has been able to make carbon nanowires yet.”