Like that little old ant with the rubber tree plant in the song, silicon wires just four atoms wide and an atom tall may be small, but they can carry a load.
Purdue Professor Gerhard Klimeck's lab helped a team from the University of New South Wales and Melbourne University in Australia demonstrate the potential of the tiny wires the Australian researchers developed by showing that the devices have the same current-carrying capability as copper wires.
Experiments and the Klimeck research group's atom-by-atom supercomputer models, developed and run on Purdue's Coates and Rossmann community clusters, indicate that the nanowires maintain a low capacity for resistance despite being more than 20 times thinner than conventional copper wiring in microprocessors.
“Having the throughput capability for a highly scalable code is important for doing this kind of simulation, and we have that capability at Purdue,” says Klimeck, director of the Network for Computational Nanotechnology and nanoHUB.org. “We ran hundreds of cases to understand the potential landscape of these devices, so this was computationally intensive work.”
The discovery, published in the journal Science, has several implications, including:
The innovation of the Australian group was to build the circuits up atom by atom, instead of the current method of building microprocessors, in which material is stripped away, says Klimeck, a Purdue professor of electrical and computer engineering.
“Typically we chip or etch material away, which can be very expensive, difficult and inaccurate,” Klimeck says. “Once you get to 20 atoms wide you have atomic fluctuations that make scaling difficult. But this experimental group built devices by placing atomically thin layers of phosphorus in silicon and found that with densely doped phosphorus wires just four atoms wide it acts like a wire that conducts just as well as metal.”
Hoon Ryu, a Purdue graduate who's now a senior researcher with the Korea Institute of Science and Technology's Supercomputing Center, said the practicality of the research is exciting.
“The metallic wire is in principle quite difficult to be scaled into one- to two-nanometer pitch, but in both experimental and modeling views, the research result is quite remarkable," Ryu says. "For the first time, this demonstrates the possibility that densely doping wire is a viable alternative for the next-generation, ultra-scale metallic interconnect in silicon chips.”
“We were doing simulations of experimental work, which was based on a theoretical model,” Klimeck says. “So we were bringing the three legs of modern science — theory, experiment and computer simulation — together in one project. Plus, our graduate students are able to stay in contact and work with each other despite working in various locations around the world. It's hard to think of a better example of how science is done today.”
“Having the throughput capability for a highly scalable code is important for doing this kind of simulation, and we have that capability at Purdue”