Lasers have become indispensable for even the most humdrum tasks, from highlighting Powerpoint presentations to burning music CDs. Lasers are also essential for high-speed communications along optical fiber, which has vastly greater bandwidth and much less crosstalk than electrical transmissions in copper wire. Recently scientists developed lasers made from silicon, an important first step in the development of high-speed chips that will fully integrate light-speed communications with the processing power of silicon electronics.
As CPU processing gets faster, the need increases for near-instantaneous clock synchronization within CPUs as well as for fast interchip communications for parallel computation. The integrated-circuit industry is rooted in silicon technology. Anything that can be made out of silicon can be fabricated with submicron dimensions and in huge volume, with great reliability. But silicon's electronic properties prevent it from functioning as a conventional laser. The material has an “indirect bandgap,” which means that electrons cannot emit photons by dropping directly from one energy level to another. Solid-state lasers have until recently been made from direct-bandgap materials such as gallium arsenide (GaAs), which can spit out photons in the desired manner. Making an interface between the GaAs devices and silicon systems is difficult, and results are hard to reproduce to industry specifications.
A technique called Raman scattering, though, has overcome this problem. The process begins when electrons first absorb photons. The excited electrons then “scatter” energy by emitting both phonons—a vibration of the silicon's lattice crystal—and photons of lower energy than the ones absorbed. In October 2004 Ozdal Boyraz and Bahram Jalali, two engineers at the University of California, Los Angeles, announced that they had demonstrated the first silicon Raman laser. It was an infrared device that emits pulses, each lasting 25 trillionths of a second, far shorter than the interval between them. The short pulses were necessary because of a two-photon absorption effect. Silicon atoms can absorb two photons simultaneously, generating an electron and a hole (the absence of an electron). The electron-hole pair remains in the material for a long time, absorbing power and weakening the laser amplification. The long gaps between the pulses in the Raman laser allow the electrons and holes to dissipate.
In February, Intel's Haisheng Rong and his colleagues published a paper in Nature detailing construction of a continuous-output silicon laser that attacked the two-photon absorption effect in a different way. Their device, a five-centimeter-long, S-shaped silicon waveguide, cleverly sidestepped the issue. Rong used a classic silicon device, the PIN diode. He doped one side of the waveguide with positive charge and the other with negative charge and then applied a voltage sideways across the waveguide to remove electron-hole pairs generated by two-photon absorption before they could absorb laser power.
Rong's innovative device exploited the same five-centimeter length of silicon for use as both an infrared laser and a semiconductor diode. His silicon laser is a significant advance because continuous beams, which can be modulated and chopped, provide the basis for digital communications. Low-cost optoelectronic devices made entirely from an industry-standard silicon process are still a long way off, but these lasers build a foundation from which we can expect to see light-speed information processing technology develop into reality.