For years, engineers and scientists have dreamed of the advantages that could be gained by integrating photonics technology and semiconductor electronic chips. Computers, cell phones and other electronic devices could shrink dramatically, while their speed could increase by orders of magnitude. But one stumbling block has stood firmly in the way of such integration: Silicon, the material of all semiconductor electronics, is an indirect- bandgap material. When free carriers -- electrons and holes -- combine in silicon, the energy is released in the form of a phonon rather than of a photon. As a result, normal LEDs and laser diodes cannot be fabricated from silicon, and the integration of photonic and semiconductor electronic chips has remained a dream.That is not to say, however, that prodigious amounts of intellectual energy have not been devoted to efforts to overcome silicon's indirect bandgap. But none of these efforts has resulted in a robust silicon laser that likely would lead to the integration of photonics and semiconductor electronics.During the past year or so, the Raman laser has emerged as one of the most promising approaches to a viable silicon laser. Late last year, a group at the University of California, Los Angeles, reported the first silicon-based Raman laser.Although this was an exciting development, it had several shortcomings. It was optically, rather than electrically, pumped, so it did not provide a single-chip solution to the silicon laser problem. Moreover, the optical pumping had to be performed with very short pulses, making it too expensive to be practical for consumer goods. Finally, it required several meters of optical fiber for its resonator, which made it too bulky for most applications.Recently, a research group from Intel Corp. in Santa Clara, Calif., and Jerusalem reported a silicon Raman laser that overcomes the last two of these three limitations.Raman lasers are fundamentally different from other lasers in that the coherent light originates from stimulated Raman scattering rather than from a population inversion of ions or atoms. In a silicon Raman laser, photons are created not by the recombination of an electron and a hole, but by scattering of the pump light from a lattice vibration. Thus, silicon's indirect bandgap is not a consideration in a Raman laser. Figure 1. The silicon Raman laser is based on a rib waveguide with dimensions W = 1.5 µm, H = 1.55 µm, h = 700 nm and D = 6 µm. A reverse bias applied across the horizontal PIN junction swept free carriers out of the waveguide and dramatically reduced optical losses. Images ©Nature Publishing Group.If silicon's indirect bandgap can so easily be finessed, why has it taken so long for silicon Raman lasers to be developed? The biggest issue is probably free-carrier absorption in silicon of both the pump light and the coherent Raman signal. The free carriers are created when two-photon absorption rips an electron from one of the silicon atoms, creating a mobile electron and a hole. Although the two-photon absorption itself is not a critical loss mechanism, the resulting free carriers can easily absorb enough photons to prevent the laser from reaching threshold.Several techniques to minimize the detrimental effects of free-carrier absorption have been proposed. The lifetime of the free carriers can be reduced, for example, by lateral scaling of the laser waveguide's modal area.Figure 2. Transmission through the silicon waveguide was increased by holding the P- and N-contacts at the same voltage (V = 0), and increased even more by applying a reverse bias (25 V). The indicated free-carrier lifetimes were calculated from the transmission data.In demonstrating the first silicon Raman laser, the University of California researchers took a different approach. They pumped the Raman laser with a mode-locked laser whose 30-ps pulses were much shorter than the free-carrier lifetime and whose period was much longer than the carrier lifetime. Under these conditions, the laser pulse came and went before there were sufficient free carriers to absorb the photons.The Intel scientists avoided the use of an expensive pulsed pump laser by integrating a PIN junction into the silicon device (Figure 1). Optical transmission through the waveguide increased dramatically when a reverse voltage was applied to the junction (Figure 2). The bottom trace in Figure 2 shows the transmission through the 4.8-cm-long waveguide when there was an open circuit between the P- and N-contacts. Although the transmission started near unity for low input power, it decreased sharply as free carriers were created by the increasing input power. Figure 3. The resonator for the Raman laser was formed by a 90 percent high-reflective coating on the back end of the S-shaped waveguide and the 30 percent Fresnel reflection from the uncoated front end. Although laser light was emitted from both ends, only the beam from the front was measured.The middle trace shows that this deleterious effect was diminished when the P- and N-contacts were connected; that is, held at the same voltage. When a reverse bias of 25 V was applied in the top trace, free electrons in the intrinsic section of the semiconductor -- in the optical waveguide -- were pulled back toward the N-doped region, and holes were pulled back toward the P-doped region. Because these free carriers had been removed from the waveguide, the transmission through the waveguide increased significantly.Figure 4. The saturation evident at the upper end of the input-output curve resulted from free carriers being produced in the silicon waveguide faster than the voltage applied across the PIN junction could sweep them out.The rib-waveguide laser was fabricated in an S-pattern on an undoped, 2 × 16-mm silicon-on-insulator substrate using standard projection lithography and plasma reactive ion etching. The Intel scientists optically pumped it through a thin-film multiplexer with an amplified, 1.54-µm, CW diode laser (Figure 3). To further mitigate residual free-carrier absorption and minimize thermal effects, they chopped the pump beam with an acousto-optic modulator to produce 130-ns square pulses at a 10-kHz repetition rate. The laser resonator was formed by a 90 percent reflective coating on the back end of the waveguide and the 30 percent Fresnel reflection from the uncoated front end.Figure 5. Without the resonator mirrors, the spontaneous Raman scattering was much broader than the stimulated Raman scattering observed in the resonator. Note that the spontaneous trace is amplified by a factor of 105.The laser output, observed from the front end of the waveguide, indicated a laser threshold of 0.4 mW and a slope efficiency of 9.4 percent (Figure 4). The scientists estimate that the slope efficiency would have been approximately 10 percent had they included the output from the back of the resonator. Spectral narrowing in the presence of a resonator is a strong indication of laser action (Figure 5).