Ask today's physics graduates if silicon can produce stimulated emission and show optical gain, and most would say no. But students graduating in the next few years will enter into an optoelectronics industry that has achieved the seemingly impossible: a silicon laser. Silicon microphotonics has boomed in the last few years. Many silicon-based devices have been demonstrated: waveguides, tunable optical filters, fast switches, fast optical modulators, fast CMOS photodetectors, photonic crystals and microelectromechanical systems. But not a laser. A silicon laser would, for the first time, allow monolithic integration of photonics and electronics to share the same chip. Despite huge research efforts by groups around the world, a silicon laser — the device that promises to revolutionize the optoelectronics industry — remains elusive. However, many researchers have made silicon-based LEDs and a few have shown gain. The most interesting aspect of silicon laser research is that no two teams are developing the same technology. Each is trying different and novel approaches in an attempt to squeeze light out of silicon. Groups at the Università di Trento in Italy and at the University of Rochester in New York are investigating silicon nanocrystals and are the only ones to have reported gain; researchers at the University of New South Wales in Sydney, Australia, have made an efficient LED using bulk silicon and lessons learned from the photovoltaics industry; scientists at the University of Neuchâtel in Switzerland are developing silicon-based quantum cascade lasers; a group at Surrey University in the UK fired boron atoms at bulk silicon to engineer local dislocations and make a relatively efficient LED; at STMicroelectronics in Sicily, the only group to be backed by a large electronics manufacturer is using silicon-rich oxide as a host for active ions; and another team at the University of New South Wales is using porous silicon for the same purpose. A laser host Porous silicon was at one time thought to be the answer to the quest for a silicon laser, but after several years of work, researchers have been unable to increase the efficiency of light emission. This is why Mike Gal’s group at the University of New South Wales is using porous silicon not as a light emitter, but as the laser host that contains the optically active ions. “We have already made a number of very high quality porous silicon microcavities doped with erbium using ion implantation. With better cavities, we think we will be able to get lasing in doped porous silicon microcavities,” he said. “We think this will be a viable method to make silicon lasers because, while we can make excellent optical microcavities out of porous silicon, it is not an efficient emitter of light.” And emission efficiency is the first milestone in the race to develop a silicon laser. The LED devices made by Salvo Coffa’s group at STMicroelectronics hold the world record for efficiency. He and his colleagues used ions of rare-earth metals such as erbium or cerium implanted in a layer of silicon-rich oxide — silicon dioxide enriched with silicon nanocrystals 1 to 2 nm in diameter (Figure 1). Figure 1. The LED devices made by Salvo Coffa’s group at STMicroelectronics hold the world record for efficiency. His colleagues used ions of rare-earth metals such as erbium or cerium implanted in a layer of silicon-rich oxide. “Our device has an internal quantum efficiency of 50 percent and an external efficiency of 10 percent,” he said. The frequency of the emitted light depends on the choice of rare-earth dopant — 1.54 μm for erbium — and STMicroelectronics has patented techniques for implanting the rare-earth ions in the silicon. “It is important to get the correct density of active ions and to ensure they do not cluster, or the optical properties will be lost,” Coffa said. “It is also important to get a uniform electric field across the device to prevent hot spots.” The company has also patented a structure in which two circuits, built on the same chip but electrically separated by insulating silicon dioxide, communicate via optical signals using integrated silicon light emitters and detectors. These devices will have numerous applications, including motor control, power supplies, solid-state relays and others where the power circuit must handle much higher voltages than the control circuit does. In the longer term, the company is investigating optical data transmission systems as well as low-cost integrated devices for dense wavelength division multiplexing. Although many applications use an LED as a light source, silicon LEDs have slow switching times and some applications need faster modulation. Coffa is confident that building a laser is the solution. “We hope to have an optically pumped laser by July 2003 and an electrically pumped laser by the end of 2003.” Before Coffa’s work, the efficiency world record for a silicon LED was held by Martin Green’s group at the University of New South Wales.1 The group’s expertise is in developing photovoltaic cells, not LEDs. “This is a classic example of learning from work in a related area,” Coffa said. “Green and his colleagues have optimized absorption processes and used ultrapure silicon. This means that, because of the absence of defects, the carriers have no place to recombine and will eventually emit a photon.” Green’s electrically pumped LED uses bulk silicon and emits at 1.15 μm near room temperature (Figure 2). It has an internal quantum efficiency of 10 percent and an external efficiency of 1 percent — approximately the same levels obtained from GaAs devices 10 years ago. Figure 2. At the University of New South Wales, Martin Green’s electrically pumped LED uses bulk silicon and emits at 1.15 μm at near room temperature. These efficiencies were obtained by drastically reducing nonradiative recombinations. “We used small contact regions, low-doped and extremely pure material, and the light extraction efficiency was enhanced by texturing the surface,” Green said. High emissivity is obtained using inverted pyramids on the top surface, formed by anisotropic etching. These pyramids not only reduce reflection, but, more importantly, also increase emissivity by trapping weakly absorbed light within the cell. Observers of his work say that the need for ultrapure silicon can be a disadvantage. But Green disagrees. “The need for high-quality silicon is not such a big problem,” he said. “We don’t work in a cleanroom, so I think the industry is capable of obtaining the same quality standards that we have achieved.” But he does not believe bulk silicon will yield a laser. “I don’t think we will be able to get lasing in bulk silicon because we need very high carrier densities. This is why we are going to make 2-D structures. Due to quantum confinement effects, the bandgap increases and emission at short wavelengths will occur. We expect to get photoluminescence at 0.6 μm.” The university has been granted a patent for Green’s device and has provided funding for a high-throughput testing laboratory. Using this approach, he hopes to have developed a fully functioning optically pumped silicon laser in five years and an electrically pumped laser in eight. And all this promising work has stemmed from research into photovoltaic cells. Something to work with “For many years the solar cell industry has been taking advantage of the findings of the intense research done by the world’s microchip industry to lower the cost and improve the efficiency of solar cells,” Green said. “Even though we have been seeking quite different results to those wanted by the microchip industry, we are all working with silicon. Now we have turned the tables and given the microchip industry something to work with.” He said he was spurred to investigate developing an LED when he saw a paper published by Kevin Homewood’s group at Surrey University in the UK.2 The scientists claimed an external efficiency of 0.2 percent by using an approach very different from Green’s — introducing localized defects that produce localized dislocation loops. This localization increases the recombination rate of injected carriers. The resultant LED has a tunable wavelength between 1.1 and 1.7 μm at room temperature. Russell Gwilliam, a member of Homewood’s group, said the main advantage of the approach is that “the technology is 100 percent CMOS-compatible and uses current fab tools in near standard process windows.” He acknowledged, however, that there is still much work to do to increase the efficiency and make a laser. “As the light is generated within the silicon, the external efficiency is reduced compared with oxide-based systems, due to the relative refractive index of silicon compared with silicon dioxide,” he said. “However, as a laser requires high internal reflection, this may well work in our favor.” Gwilliam is confident that his approach will lead to a silicon laser. He has started a company, Si Light Technologies, to commercialize the work and is looking for funding, he said. Of the other approaches, he is most skeptical about silicon nanocrystals. “We investigated silicon nanocrystals some years ago. We even patented the approach. But we could not get sensible charge injection into the nanocrystals because they are based on silicon dioxide. Our new silicon-based approach does not have these problems.” Lorenzo Pavesi from Università di Trento, whose work is based on silicon nanocrystals, acknowledged that charge injection may be a problem. “But we have demonstrated that silicon can show optical gain, whereas the Surrey group has not yet been able to do this,” he said. Like Homewood and Gwilliam, Pavesi published his work in Nature and raised the awareness of silicon-based LED and laser research.3 Pavesi’s optically pumped device emits from 750 to 900 nm, and some of the devices made by the group have an external efficiency of 1 percent (Figure 3). Figure 3. Lorenzo Pavesi at the Università di Trento in Italy has demonstrated that silicon can show optical gain. “There are still materials issues to be faced,” he said. “Electrical pumping may be a problem.” Philippe Fauchet at the University of Rochester agrees. His group has made silicon LEDs by layering amorphous silicon with silicon dioxide and using an annealing process to turn the amorphous silicon into an array of silicon nanocrystals. “The nanocrystal technique has one inherent problem,” he said. “For it to work, we need to fabricate isolated quantum dots, but this in turn means current injection will be poor, because the nanocrystals are isolated from each other.” His group has shown “large gain” and plans to present its findings at a materials conference that will be held in April. Silicon nanocrystals emit visible light at room temperature, and the emission band depends on the mean size of the nanocrystals: The emission band shows a blue shift and a narrowing as their size decreases. The mechanism for this luminescence is still being debated, but Pavesi believes that the emission originates in nanocrystals that are coated with a stressed silica shell. This enhances the formation of interface oxygen-related states on the surface of the silicon nanocrystals, which emit when excited. Although this approach has been successful for Pavesi, he said that not all the physics behind the technology is completely understood. Figure 4. A transmission electron microscope captured five periods of Jérôme Faist group’s quantum cascade silicon-germanium LED. The light gray lines are the Si barriers, the dark gray lines are SiGe (80 percent Ge) and, where the scale is indicated, SiGe (50 percent Ge). The small dots are atoms. The thinnest Si barrier is only 4 Å (0.4 nm) thick. This shows the group’s ability to grow SiGe layers with any Ge content with an extreme control on the layer thickness. Jérôme Faist at the University of Neuchâtel, however, has the opposite problem. He has a system — the quantum cascade laser — that is well-understood and that theoretically will produce a laser with very high efficiencies (Figures 4 and 5). But, practically, his group still has a long way to go before it demonstrates a laser. The Neuchâtel device emits at 7 μm but works only at low temperatures. “We use very thin alternating layers of silicon and germanium to confine carriers in a sandwich of layers,” Faist explained. “They cascade through these layers, emitting a photon at each step, like an electronic waterfall.” Figure 5. A scanning electron microscope captured this image of Jérôme Faist’s group’s LED. The white line is the waveguide, and along the waveguide are the top metal electrodes (dark gray). The light gray rectangles are the bottom electrodes. Most of the black surface is a silicon-nitride insulating layer. The group recently published a paper showing electroluminescence from quantum cascade structures.4 In previous structures, strain between the silicon and germanium layers was a big issue. But the scientists have used strain compensation techniques on samples consisting of up to 50 periods (about 700 quantum wells) and redesigned the active region. Using these devices, Faist believes population inversion may be possible. Enough power? Population inversion is the next milestone in the race to develop a silicon laser. But although all the aforementioned researchers believe that population inversion — and therefore a laser — will be possible, Coffa believes that there is one parameter that many have not considered. “I am positive a silicon laser will be demonstrated soon, and, obviously, I hope it is by my group, but the important question is whether such a laser will emit enough power to be useful,” he said. “For an integrated application with a very sensitive detector, nanowatts of power are sufficient. But for most other applications, milliwatts of power are needed, and I am not sure if this can be achieved by any of the groups working on a silicon laser.” There are still many obstacles to overcome: Output power, efficiencies, optical gain and radiative recombination rates need to be increased if the silicon laser is to become a reality. But if one takes into account that modern gas lasers operate with very low gain and modern solid-state lasers have low radiation decay rates, the prospect of a visible silicon laser becomes more realistic. References 1. M. Green et al. (August 2001). Efficient silicon light-emitting diodes. NATURE, p. 805. 2. K. Homewood et al. (March 2001). An efficient room-temperature silicon-based light-emitting diode. NATURE, p. 190. 3. L. Pavesi et al. (Nov. 23, 2000). Optical gain in silicon nanocrystals. NATURE, p. 440. 4. L. Diehl et al. (December 2002). Electroluminescence from strain-compensated Si0.2Ge0.8/Si quantum-cascade structures based on a bound-to-continuum transition. APPLIED PHYSICS LETTERS, p. 4700. Why Is Silicon an Inefficient Light Emitter? Silicon is an indirect bandgap material. Light emission is via a phonon-mediated process with low probabilities: Spontaneous recombination lifetimes are in the millisecond range. In standard bulk silicon, competitive nonradiative recombination rates are much higher than the radiative ones, and more of the excited electron-hole pairs recombine nonradiatively. This yields very low internal quantum efficiencies for bulk silicon luminescence. In addition, fast nonradiative processes such as Auger or free-carrier adsorption prevent population inversion for silicon optical transitions at the high pumping rates needed to achieve optical amplification.