A high-speed match made in silicon.

A high-speed match made in silicon

Putting gallium arsenide semiconductor circuits atopa silicon base is a bit like mating a Ferrari with a Honda. The components seem incompatible, but if the match were towork, the result would be an attractive combination of high performance and economy.

With gallium arsenide and silicon, such hybrid integrated-circuit chips may now be possible. Researchers at the University of Illinois at Urbana-Champaign have discovered a way to deposit gallium arsenide layers on top of silicon wafers without spreading crystal defects that ruin the electronic properties of the materials.

Until now, silicon and gallium arsenide technologies have developed somewhat independently. Gallium arsenide is useful because electrons travel about five times faster in this semiconductor than they do in silicon. Gallium arsenide also emits light, allowing it to be used for lasers or light-emitting diodes. However, the material is brittle and difficult to grow into large, defect-free crystals.

Large silicon crystals, on the other hand, are relatively easy to produce. Silicon is a better heat conductor, and more transistors and other devices can be packed into a given surface area. The cost of producing silicon chips is also significantly lower.

"Both technologies have a lot of things to offer," says electrical engineer Hadis Morkoc, leader of Illinois group. "Now we don't have to choose between silicon and gallium arsenide technology because we can have the best of both on the same chip."

The trick is to find a way of aligning the silicon and gallium arsenide crystal lattices. Normally, the structures don't quite match. For a row of 25 silicon atoms, only 24 atoms from a gallium arsenide layer are needed tofill the same space. This produces a large number of defects where the two lattices meet.

The mismatch can be overcome if the silicon base is slightly titled, says Morkoc. A gentle slope of about 4 [deg.] provides, at the atomic level, tiny steps that take care of the problem. If these steps have the right orientation with respect to the silicon crystal lattice, then the inherent bumpiness of the slope doesn't produce dislocations that thread their way into the gallium arsenide layer.

"The orientation is the key," says Morkoc. For a square silicon chip with an upper surface nearly parallel to a face of the crystal lattice, the slope rises from its low point at one corner to its peak at the diagonally opposite corner. The Illinois group had applied for a patent that covers specifications for the appropriate slope orientation.

Says George W. Turner, who is doing similar work at the Massachusetts Institute of Technology, "The next important milestone, which should silence some of the skeptics who still think this is a cute idea but will never lead to anything practical, is to demonstrate a room-temperature, continuously operating laser." Using materials containing more defects than those now available at Illinois, the MIT group and another in Japan have already produced pulsed lasers.

The combination of light-emitting gallium arsenide chips with complex, tightly packed silicon circuits would allow the development of "optical interconnects," says Turner. In some cases, far more power already goes into driving the wires that connect chips than in running the complicated silicon circuits themselves. With composite chips, the wires connecting one silicon device to another could be replaced by an efficient optical system, perhaps using optical fibers.

Morkoc is more interested in developing high-speed electronic devices. His team has already built several types of gallium arsenide transistors on silicon bases. Because all parts of an integrated circuit do not need to be equally fast, eventually it may be possible to deposit gallium arsenide at only the points on a silicon circuit where the chip must operate quickly. Says Morkoc, "That would make life a little easier."

The Illinois discovery will probably accelerate the pace of composite-chip research. Turner predicts that continuous lasers and optical interconnects may be developed within a year. More and more research groups are entering the field, and several small companies have been established to develop the technology.