Strange Bedfellows

Marriage of silicon and previously incompatible semiconductors is consummated

SILICON IS THE STUFF OF MEMORIES and microprocessors. III-V semiconductors—compounds made of elements inhabiting the third and fifth columns of the periodic table, like gallium arsenide—are the stuff of high-frequency communications chips, LEDs, and solid-state lasers. But these two types of materials have never been able to live together on the same chip. Now, researchers in Europe have found that at the nanometer scale, they can get along just fine. The researchers grew indium-phosphide and gallium-phosphide nanowires on a silicon substrate, clearing the path for the manufacture of cheaper high-frequency chips and silicon devices embedding LEDs and lasers.

A typical chip integrating these materials might consist of a silicon substrate with a large number of vertical nanowires, says Jorden van Dam of the Kavli Institute of Nanoscience at Delft University of Technology, in the Netherlands. The nano-wires could contain p-n junctions and could be embedded in an insulating polymer layer. Then, passing a current through the nanowires would cause emission of light from close to the junctions, making the device function as a tiny LED. Such nanowires could form a pixel in a display.

The problem that had stood in the way of that vision was lattice mismatch: the atoms in a III-V semiconductor are spaced slightly differently—by a few percent—than the atoms in silicon are. This misalignment, resulting in stresses in the interface layer, foiled previous attempts to grow III-V materials directly on silicon.

The Dutch team, in Delft, and a Swedish team independently came up with a quite simple idea to resolve the lattice mismatch. They showed that it is possible to grow III-V nanowires on silicon substrates, because the nanowires have a footprint so small that it doesn't generate the stresses in the interface region that arise with other kinds of materials. Both groups grew the nanowires from gold particles deposited on a silicon substrate.

The Swedish group, led by Lars Samuelson at Lund University, started by depositing an aerosol of gold particles. The Dutch team, comprising researchers from the Philips Research Laboratories in Eindhoven and the Kavli Institute, and led by Erik Bakkers of Philips, obtained their particles by heating a gold film 0.2 nm thick on the substrate until it melted into droplets. Initially, they deposited the drop-lets on germanium. Now they have put them on silicon as well [see photo, " White Spikes"].

Having obtained a substrate dotted with gold nanoparticles, each team then grew the nanowires by introducing its device into an atmosphere containing III-V semiconductor vapor at high temperature and pressure. The semiconductor material diffuses into a gold droplet, and a semiconductor nanowire starts to grow. A wire 20 nm thick can reach a length of 20 micrometers, long enough to function as an optical waveguide, for example, says Delft's van Dam. The growth is epitaxial; that is, important features of the crystal structure of the silicon or germanium substrate, such as the orientation of the lattice planes, are replicated in the nanowire. This ensures that electrical contact is maintained between the nanowire and the substrate.

Peidong Yang of the University of California, Berkeley—a chemist who has been investigating III-V nanowires for the past five years—says that this advance will markedly accelerate the development of new optoelectronic devices. Lincoln Lauhon, a materials scientist at Northwestern University, in Evanston, Ill., agrees, pointing out that the development "gives you a great deal of flexibility. You can mix materials with different functionalities."

Still, the road to commercialization of viable devices will have its share of potholes and sharp curves. Bakkers expects it will be at least two years before any devices reach the market. High on his wish list: a III-V nanowire that forms the channel of a transistor in which the gate is placed around the wire—a gate-around transistor, suitable for high-frequency applications above 150 gigahertz. "This is the ideal configuration: a channel with high mobility of electrons and a gate with very good capacitive coupling, resulting in high transconductance [current-carrying capability]," he says.

The results from both research groups clearly show that hybrid microcircuits, incorporating the desirable properties of the III-V compounds with those of cheap and ubiquitous silicon substrates, might soon find an important niche in electronics after all.

Related Stories