The enormous progress of communication technologies in the last years has increased the demand for efficient and low-cost optoelectronic functions. For several present and future applications, photonic materials—in which light can be generated, guided, modulated, amplified, and detected—need to be integrated with standard electronic circuits in order to combine the information-processing capabilities of electronics data transfer and the speed of light. Long-distance communications, local-area-networks data transfer, and chip-to-chip or even intrachip optical communications all require the development of efficient optical functions and their integration with state-of-the-art electronic functions. Silicon is the material of choice for reliable and low-cost optoelectronic integrated circuits because it is the leading semiconductor in the electronic arena and since a wellestablished processing technology exists for this material. However Si is characterized by an indirect bandgap and by a weak electro-optic effect. It is therefore not suitable for the implementation of fundamental optical functions such as light emission and modulation. At the moment, hybrid integration of compound-semiconductor optical functions with Si electronic functions is providing the gateway from electronic to photonic technology. However several strategies are being considered to engineer the optical functions of Si and to realize fully Si-based or at least Si-compatible optoelectronics.

Silicon-based optoelectronics is a diversified technology that has grown steadily but not exponentially over the past decade. Some applications—such as smart-pixel signal processing and chip-to-chip optical interconnects—have enjoyed impressive growth, whereas other applications have remained quiescent. A few important applications such as optical diagnosis of leaky metal-oxide-semiconductor-field-effect-transistor circuits, have appeared suddenly. Over the years, research and development has unveiled some unique and significant aspects of Si-based optoelectronics. The main limitation of this technology is the lack of practical silicon light sources—Si lasers and efficient Si light-emitting devices (LEDs)—though investigators are “getting close” to the LED.

Silicon-based optoelectronics refers to the integration of photonic and electronic components on a Si chip or wafer. The photonics adds value to the electronics, and the electronics offers low-cost mass-production benefits. The electronics includes complementary-metal-oxide semiconductors (CMOS), very large-scale integration (VLSI), bipolar CMOS, SiGe/Si heterojunction bipolar transistors, and heterostructure field-effect transistors. In this discussion, we will use a loose definition of optoelectronics that includes photonic and optoelectronic integrated circuits (PICs and OEICs), Si optical benches, and micro-optoelectromechanical (MOEM) platforms. Optoelectronic chips and platforms are subsystems of computer systems, communication networks, etc. Silicon substrates feature a superior native oxide, in addition to excellent thermal, mechanical, and economic properties. Silicon wafers “shine” as substrates for PICs and OEICs.

The achievement of efficient room-temperature light emission from crystalline Si is a crucial step toward the achievement of fully Si-based optoelec-tronics. However Si, the leading semiconductor in microelectronic applications, is unable to perform as well in the optical arena. In fact due to its indirect bandgap, Si does not exhibit efficient light emission and has been considered unsuitable for optoelectronic applications. Several efforts have been dedicated to overcoming this limitation. Among them, luminescence through the incorporation of rare-earth impurities has been considered In particular, erbium doping has been demonstrated as a valid approach toward achievement of efficient light emission from Si.1−43 Erbium is a rare-earth ion that, in its 3+ state, can emit photons at 1.54 μm because of an intra-4f shell transition between the first excited state (4I13/2) and the ground state (4I15/2). This emission is particularly attractive because its wavelength falls inside a window of maximum transmission for the silica optical fibers. When Er ions are inserted within a Si matrix, the excitation (4I15/2 → 4I13/2) can be achieved through the carriers provided by the host, whereas the subsequent deexcitation (4I13/2 → 4I15/2) can result in a sharp, atomlike light emission.

Light emission in silicon has been intensively investigated since the 1950s when crystalline silicon (c-Si) was recognized as the dominant material in microelectronics. Silicon is an indirect-bandgap semiconductor and momentum conservation requires phonon assistance in radiative electron-hole recombination (Figure 1a, top left). Because phonons carry a momentum and an energy, the typical signature of phonon-assisted recombination is several peaks in the photoluminescence (PL) spectra at low temperature. These PL peaks are called “phonon replicas.” High-purity c-Si PL is caused by free-exciton self-annihilation with the exciton binding energy of ~11 meV. The TO-phonon contribution in conservation processes is most significant, and the main PL peak (~1.1 eV) is shifted from the bandgap value (~1.17 eV) by ~70 meV—that is, the exciton binding energy plus TO-phonon energy (Figure 1a).

The need for integrated optical interconnects in electronic systems is derivedfrom the cost and performance of electronic systems. If we examine the cost of all interconnects, it becomes apparent that there is an exponential growth in cost per interconnect with the length of the interconnect. A remarkable feature of interconnect cost is that the exponential relation holds over all length scales—from the shortest interconnects on a chip to the longest interconnects in global telecommunications networks. Longer interconnects are drastically more expensive, and these costs are ultimately related to the labor cost associated with each interconnect. Given this economic pressure, it is not surprising that there is a driving force to condense more functions locally on the same chip, board, or system. In condensing these functions, the number of long interconnects are decreased and the overall cost of the electronic system decreases dramatically. A specific glaring example of this driving force is Si complementary-metal-oxide-semiconductor (CMOS) technology, especially the case of microprocessors. In the Si microprocessor case, the flood gates to interconnect condensation were opened and the miraculous trend of lower cost for exponentially increasing performance was revealed.

Thin-film integrated optics is becoming more and more important in optical-communications technology. The fabrication of passive devices such as planar optical waveguides, splitters, and multiplexers is now quite well-developed. Devices based on this technology are now commercially available. One step to further improve this technology is to develop optical amplifiers that can be integrated with these devices. Such amplifiers can compensate for the losses in splitters or other optical components, and can also serve as pre-amplifiers for active devices such as detectors.

In optical-fiber technology, erbium-doped fiber amplifiers, are used in long-distance fiber-communications links. They use an optical transition in Er3+ at a wavelength of 1.54 μm for signal amplification, and their success has set a standard of optical communication at this wavelength. Using the same concept of Er doping, planar-waveguide amplifiers are now being developed. For these devices, silicon is often used as a substrate, so that optoelectronic integration with other devices in or on Si (electrical devices, or Si-based light sources, detectors, and modulators) may become possible. Figure 1 shows an example of a silicon-based optical integrated circuit5 in which a 1 × 4 splitter is combined with an amplifying section.

Photodetectors must provide fast and efficient conversion of photons to charge carriers. When considering semiconductor light sources, the indirect bandgap of silicon and germanium represents a serious obstacle to radiative electron-hole recombinations. Momentum conservation demands the simultaneous interaction of the electron-hole pair with a momentum-matching phonon. As a consequence, radiative recombinations are five orders of magnitude less probable in Si if compared to a direct semiconductor such as GaAs.

Although the absorption of a photon and the generation of an electron-hole pair may be considered as the inverse process to emission, photon absorption within indirect semiconductors is a highly probable process if the photon energy is sufficient to bridge the energy gap in a direct process. The resulting electronhole pair is created in an excited state and relaxes sequentially. The ubiquitous-silicon solar cells operate this way. In the visible spectral range, Si photodetectors have demonstrated fast and efficient performance, being readily adapted for opto electronic applications and being fully compatible to standard-silicon processing schemes.

Over the past decade or so, research in silicon-based heterostructures has evolved from a few seminal publications on the growth and physical properties of Si1−xGex heteroepitaxial layers to a technology currently entering large-scale commercial production for heterojunction bipolar transistors (HBTs). During this period, extensive work has taken place on the optoelectronic applications of Si/Si1−x Gex such as 1.3–1.55 μam detectors for optical communication, 2–12-μm infrared detectors for two-dimensional (2D) focal plane arrays for night vision and thermal imaging, and infrared emitters for chip-to-chip optical communication as well as waveguiding and modulators. The overall goal of this work has been to merge optoelectronic functionality with the very large-scale-integration and electronic signal processing capabilities of silicon to create a silicon-based “superchip.”