Silicon devices: LED there be light.

Silicon devices: LED there be light

Silicon is king of the electronics world. Virtuallyevery kind of semiconductor device can be made from this material. The one exception has been a device that emits light. For example, while silicon has enabled all sorts of calculations to be performed on hand-held calculators, only gallium arsenide -- which makes up the light-emitting diodes (LEDs) in the calculator's display -- has allowed the numeric inputs and outputs to be illuminated.

But if papers presented April 22 at theAnaheim, Calif., meeting of the Materials Research Society are any indication, silicon's exclusion from the light-emitting club may be soon challenged.

Researchers from AT&T Bell Laboratorieshave developed a technique enabling them to construct novel artificial crystals that could soon lead to the first silicon-based light-emitting devices. If they succeed, says Kevin J. Malloy, a program manager for electronic materials at the Air Force's Office of Scientific Research in Washington, D.C., who attended the meeting, "they will open up a whole host of optical and electronic applications that were previously reserved for gallium arsenide."

What's more, because their techniquegives the researchers remarkable control over the small-scale structure of crystals, it will allow them to custom-tailor the properties of a great variety of materials in addition to silicon, and to create crystals that have never before appeared in nature.

Gallium arsenide naturally emits lightbecause it has what is known as a direct-energy band gap: When free, conducting electrons drop to a lower-energy, bound state, they transfer their energy directly to light. In contrast, the band gap of bulk silicon is indirect: The energy given up by electrons is largely converted into phonos, or vibrations of the crystal lattice.

The AT&T researchers have discoveredhow to change the electronic structure of silicon in a way that may allow optical transitions to occur more easily in the material. They use a conventional technique called molecular beam epitaxy, but they grow their crystals in an unusually high vacuum and at low temperatures, and they are able to meter out, atom by atom, the material that gets deposited on the crystal. This gives them a great deal of control over the placement of the atomic building blocks, so that they can construct crystals one atomic layer at a time. The growth and study of these crystals have been conducted by Joze Bevk, Thomas P. Pearsall, Leonard Feldman and John Bean at Murray Hill, N.J., Abbas Ourmazd at Holmdel, N.J., and their colleagues.

By alternately depositing layers of siliconand germanium atop a thicker silicon substrate, they have constructed crystalline films in which the atomic constituents vary over the unusually small distances of one or two atomic layers. "When crystals are grown with most conventional techniques, it's hard to grow anything thinner than a few hundred atomic layers," says Pearsall.

His group can make crystals severallayers thick that consist of, for example, alternating rows of two atomic layers of germanium and two of silicon. At this small scale, the researchers are altering the structure of the unit cell, which is the smallest collection of atoms that contains all the properties of the bulk material; normal unit cells for silicon and germanium are four atomic layers high.

The properties of the new crystals,says Pearsall, are different from those of bulk silicon, bulk germanium and germanium-silicon alloys made by mixing the two elements together.

Pearsall says his group has experimentallydetermined that the artificial ordering of atoms has produced new, strong optical transitions that lie 0.7 to 2.0 electron-volts (eV) above the bound state. This does not mean, however, that the crystals will emit light: There is an indirect energy transition lying 0.1 eV below the lowest optical transition, and because this indirect transition occurs at a lower energy, it is more likely that electrons would lose their energy to phonons than to light. Still, this is much closer to being a direct-band-gap material than the bulk materials; the lowest direct transition in a 50 percent silicon, 50 percent germanium alloy, for example, occurs at 2.6 eV.

With some guidance from SverreFroyen at the Solar Energy Research Institution in Golden, Colo., and his colleagues (who presented the results of their theoretical calculations on the new materials at the Anaheim meeting), the researchers think they may be able to get even closer. Bevk says the key factors for creating new energy levels and for controlling where the indirect band gap lies are the order, periodicity and mount of strain in the crystal on the atomic level.

AT&T's materials are strained becausethe lattice constant -- the natural distance between atoms in a crystal -- of silicon is smaller than that of germanium. Since the silicon substrate sets the lattice constant for the rest of the crystal, the germanium atoms are squeezed together closer than they would be naturally in a pure germanium crystal. Unfortunately, says Pearsall, this compressive strain on the germanium causes the indirect band gap to be the lowest-energy feature. "If we could make the strain tensile instead of compressive," he says, "it would force this indirect band gap to go up in energy, greatly improving the changes of getting light out of the crystal."

Pearsall and Froyen think they can dothis by using different substrates, such as germanium or a silicon-germanium alloy. In these cases, the silicon atoms would then be put under tensile strain. "Producing these samples is probably not going to be any more difficult than producing the samples we've worked on so far," says Bevk.

If he and his colleagues succeed, theywill have found a way to cut down on the costs and times of producing light-emitting devices. "By processing just one 6-inch silicon wafer alone," Pearsall says, "I could make enough LEDs to satisfy all the needs of [AT&T] for probably four or five years."

Gallium arsenide is still more of alaboratory technology than a production technology, he says. It would take about a week to make 20 to 30 working LEDs from one gallium arsenide wafer; with one 6-inch silicon wafer, over 1 million devices could be produced in a day, he notes.

Having silicon light devices would alsocut down tremendously on the costs and time involved in connecting gallium arsenide LEDs to silicon integrated circuits. These assembly costs, Pearsall notes, are "very expensive. You'd really like to have everything on one little chip of silicon."

In addition to making silicon LEDs, theresearchers envision a variety of other uses, including nonlinear optical devices for handling light in fiber optics and in making silicon solid-state lasers. Since their technique is not limited to silicon and germanium, many other potential applications are possible, they say, such as making better superconductors.

"I would expect that in every casewhen you take a material that is homogeneous and change it into something that has a dramatically varying atomic or chemical potential over only a few angstroms, you will see spectacular properties because you'll be exaggerating the normal forces that exist in most materials," says Pearsall.

But for now most of these ideas mustwait until the researchers' understanding of the technique and its implications develops more fully. "In a way we're like children now," says Ourmazd. "We've discovered a new game and we're trying to learn the rules."