In the past three years, several research groups have demonstrated Raman lasers in silicon. That is important because silicon is the basic building block of semiconductor electronics, and the integration of electronics and photonics would be greatly enhanced by the existence of a silicon laser. Silicon’s indirect bandgap prevents it from lasing as other semiconductors do — when holes and electrons combine — and researchers have finessed this problem by creating a laser based on a completely different physical phenomenon, stimulated Raman scattering. Figure 1. The ring-laser cavity, fabricated as a silicon-on-insulator rib waveguide, was pumped at 1550 nm and generated Raman laser output at 1686 nm. Images reprinted with permission from Nature Photonics. These early Raman lasers, although definitely a scientific breakthrough, lacked the requirements for practical applications such as telecommunications and remote sensing. Many were pulsed rather than continuous, had high laser thresholds requiring high pump powers, and needed an external voltage to operate. Recently, Intel scientists in Santa Clara, Calif., and in Park Har Hotzvim, Israel, demonstrated the first silicon Raman laser that avoids these problems.The scientists configured the silicon resonator as a ring and pumped it with 1550-nm light from a continuous-wave diode laser (Figure 1). The output Raman laser light, at 1686 nm, was separated from the incoming pump light with a wavelength filter and sent to a power meter and an optical spectrum analyzer. Figure 2. By implanting boron (red) and phosphorus (light blue) on opposite sides of the waveguide (dark blue), the scientists created a p-i-n junction across the waveguide, using it to sweep free carriers out of the waveguide.Two-photon absorption is a major problem for photonic devices in silicon. When a pair of near-infrared photons are absorbed by the material, they liberate free carriers — one electron and one hole. The presence of these free carriers dramatically increases the optical loss and, until now, had required either that the laser operate in a short-pulse mode (so the laser pulse has come and gone before the free carriers have built up) or that the free carriers be swept out of the optical region with an external voltage.Figure 3. The transmission through a p-i-n waveguide increased with increasing reverse bias across the junction. The increased transmission indicates a decrease in free-carrier lifetime from 15.3 to 0.32 ns. In this plot, the squares are experimental data points, and the lines are calculated curves for the indicated free-carrier lifetime.The scientists used the second approach, fabricating a p-i-n junction across the waveguide that formed the ring resonator (Figure 2). To demonstrate the effectiveness of this technique, they measured the throughput of a 4.6-cm-long waveguide that was fabricated on the same silicon chip as the ring cavity and, as with the ring cavity, had a p-i-n junction across it. By applying a reverse voltage across the junction, they saw that the free-carrier lifetime in the waveguide decreased from 15.3 to 0.32 ns (Figure 3).A similar reduction in free-carrier lifetime was apparent in the laser itself (Figure 4). With the optimal bias voltage applied, it began lasing at a pump power of 20 mW and generated up to 50 mW of output with a slope efficiency of 28 percent. The 20-mW threshold is an order-of-magnitude improvement over earlier Raman lasers in silicon, and the output power and slope efficiency represent an improvement factor of five.Figure 4. With a 25-V reverse bias across the p-i-n junction, the Raman laser generated up to 50 mW at the Stokes wavelength. Significantly, the laser reached threshold and produced output even with zero voltage across the junction. The cavity length (ring circumference) for these data was 3 cm, but the scientists obtained similar results in a 1.5-cm cavity, indicating that scaling to smaller sizes is feasible in photonic integrated circuits.As the scientists reduced the voltage across the junction, the output power diminished and showed signs of saturation as a function of input power (Figure 4). The threshold remained at about 20 mW, however, because two-photon absorption produces a negligible number of free carriers at optical powers under 20 mW. But even with a zero-voltage bias across the junction, the laser still reached threshold and lased. This is the first demonstration of continuous-wave lasing in silicon without an external bias. The ability to operate in an all-optical (i.e., no electricity required) manner is important in environments such as remote sensing, where electricity is not accessible.Nature Photonics, April 2007, pp. 232-237.