Nanotech reality check: Engineers take the temperature of new transistor designs
Nanoscience and nanotechnology excite researchers because to delve into the scale of molecules and atoms is to explore a strange new world, where materials and structures take on unusual and potentially useful properties. Different as it is, this world is nevertheless real. It forces trade–offs and demands compromises. Mechanical engineering Associate Professor Kenneth Goodson and electrical engineering Research Associate Eric Pop, can tell you that while the menu has changed, there are still no free lunches.
Many engineers, for example, are cleverly exploiting nanotechnology to make smaller and smaller transistors, the basic components of all computer chips. But Goodson and Pop study how transistors generate heat. They are finding that nanotechnology often brings new problems of heat dissipation that must be resolved before these new-fangled transistors can work.
“What I’ve learned as an electrical engineer,” Pop says, “is that all the tricks electrical engineers pull to carefully design the electrostatics of these devices, to make them electrically better, often end up hurting their thermal properties.” Goodson and his former students Pop, and Sanjiv Sinha (PhD 2005 ME) detail the heat problems raised by nanotechnology in a paper slated for publication in an upcoming issue of the Proceedings of the IEEE.
Certainly Goodson and his students are not trying to discourage or gainsay their fellow engineers. By studying problems associated with heat, they can help electrical engineers account for and avoid them. “The brightest minds in EE are inventing new generations of nanotransistor technologies, however many of these will be possible only through careful thermal design,” Goodson says.
A matter of scale
The small scale that defines nanotechnology gives rise to a fundamental problem with heat. Transistors are becoming so small that they are leaving little room for the physical processes that dissipate heat to operate. Here’s what’s going on: Temperature quantifies the intensity of atomic motion in a material. When electrons move through a silicon transistor, they increase the amplitude of vibrations of the atoms within the crystal structure (like a passing truck can shake nearby windows), increasing its temperature. At first these vibrations called phonons oscillate very quickly but move very slowly, making it hard for the hotspot within the transistor to cool off. Eventually they interact with their environment, transferring their energy to faster moving “acoustic” phonons (so called because they behave much like sound waves). The acoustic phonons leave the transistor quickly and things cool off. The problem with the space crunch is that it impairs the energy transfer from slow to fast vibrations and makes it harder for the acoustic phonons to leave. This leads to much higher temperatures in the transistor.
As if decreasing volumes weren’t enough of a problem, another one is that as dimensions shrink, the ratio of surface area to volume goes up (think of shrinking the radius of a soda can down to the radius of a spaghetti strand). An increased role for surface area is a problem because phonons—which carry heat away—typically have a lot of trouble crossing boundaries from one material into another. Instead of passing through, they might instead bounce off, as sunlight is sometimes reflected by the surface of a pond. In transistors, these boundaries exist along material surfaces, so the more important surface area is, the more potential there is for heat to have trouble escaping.
Materials and connections
Material surfaces are not only more important, but also more prevalent in the newest generations of transistors. The reason for the added complexity is that the tiny scale of nanotechnology presents electrical problems that engineers must stave off with novel transistor structures and exotic materials. Some of these schemes, while successful electrically, create new heat dissipation challenges. One example is an increasingly prominent transistor technology called silicon-on-insulator (SOI). An SOI transistor has a layer of insulating material underneath that lowers its electrical capacitance. The result is that the transistor can switch on and off more quickly. But it turns out that this electrical insulation is about 100 times worse than silicon at conducting heat, Goodson says.
Despite their thermal drawbacks, SOI transistors have begun finding their way into chips. A more exotic pair of incredibly thin computer chip building blocks — silicon nanowires and carbon nanotubes — are still under development in research labs. It is not too early, however, for engineers to consider their thermal properties, Goodson says. This is because Pop has found that both the nanowires and nanotubes heat up considerably when carrying significant amounts of current. In fact, nanotubes exposed to air in the laboratory actually burn, breaking instantly when a large enough current is passed through. In a computer chip, nanotubes and nanowires would be surrounded by a layer of insulating material (like snakes in sweaters), so they wouldn’t burn up. But the insulation would still trap heat inside. While both nanotubes and nanowires might still prove extremely useful in future chips, Pop says that electrical engineers should temper some of the early exuberance about these structures, in light of these heat dissipation questions.
There is little doubt among engineers that nanoscience and nanotechnology hold tremendous potential for revolutionizing not only electronics, but also fields such as environmental science, medicine and energy. At the end of the day, however, the promise won’t be fulfilled if innovations on paper cannot work in real products. Goodson and Pop see their work as ensuring that good ideas are fully feasible. Their investigation of heat in nanoscale transistors is in that spirit. “We’ve been the first people to start looking at these things — to kind of look ahead and ask whether these things are really worth it electrically and thermally,” Pop says. “Because if they’re not, then why bother with the manufacturing?”
