The first sophisticated electronic circuits made from germanium, a promising alternative to silicon, show a path for the computer industry to keep advancing beyond the physical limitations now being reached. Researchers from Purdue University demonstrated the circuits this week at the International Electron Devices Meeting in San Francisco.
Switching from silicon to germanium would be an ironic twist. The first transistor, created at Bell Labs in 1947, was made of a slab of germanium, an element one spot below silicon in the periodic table. Germanium was tried because charge flows through it very rapidly, a key property for a transistor. But as engineers worked out how to make integrated circuits and manufacture them on a large scale, germanium was set aside for silicon because it’s easier to work with.
Now, as manufacturers see problems with the continued miniaturization of silicon, germanium is experiencing a revival. The germanium circuits demoed by Purdue University engineer Peide Ye and colleagues suggest that the material could be ready for commercialization in a few years.
The tiniest transistors in production today are just 14 nanometers across, and they’re packed incredibly closely together. The semiconductor industry is finding that scaling any smaller introduces a range of problems. At one panel held during the IEDM conference, Mark Bohr, a senior fellow at Intel, estimated that silicon scaling would end in about a decade. “My general response is wild enthusiasm for any new idea,” he said.
With superb electrical properties, germanium has always promised to make speedier circuits than silicon. But engineers were unable to use it to make compact, power-efficient circuits based on the industry’s established manufacturing technique, known as complementary metal-oxide semiconductor, or CMOS, technology.
CMOS circuits use transistors that conduct negative charges, called nFETS, and transistors that conduct positive charges, called pFETs.
“Germanium pFETs are a slam dunk,” Krishna Saraswat, an electrical engineer at Stanford University who is not involved with Ye’s project, but nFETs have been the bottleneck. Ye came up with a new design for germanium nFETs that improves their performance dramatically.
Saraswat is partly responsible for reviving interest in germanium, when, in 2002, he published the first paper describing high-performance germanium transistors, which were two to three times better than silicon equivalents. “The basic science has been done, and now we’re seeing work on the basic engineering,” Saraswat says.
Other alternative materials, such as carbon nanotubes or compound semiconductors, which are made from multiple elements, also show promise for supplanting silicon, but they will be harder for the chip industry to learn to use. In contrast, chipmakers already use germanium in silicon pFETs. “Anytime you can deal with an elemental semiconductor like silicon or germanium, it’s easier,” says Xiuling Li, an engineer at the University of Illinois at Urbana-Champaign.