412 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 12, NO. 3, MAY/JUNE 2006 Raman-Based Silicon Photonics Bahram Jalali, Fellow, IEEE, Varun Raghunathan, Student Member, IEEE, Dimitri Dimitropoulos, and ¨ Ozdal Boyraz, Member, IEEE (Invited Paper) Abstract—This paper reviews recent progress in a new branch of silicon photonics that exploits Raman scattering as a practi- cal and elegant approach for realizing active photonic devices in pure silicon. The large Raman gain in the material, enhanced by the tight optical confinement in Si/SiO2 heterostructures, has en- abled the demonstration of the first optical amplifiers and lasers in silicon. Wavelength conversion, between the technologically impor- tant wavelength bands of 1300 and 1500 nm, has also been demon- strated through Raman four wave mixing. Since carrier generation through two photon absorption is omnipresent in semiconductors, carrier lifetime is the single most important parameter affecting the performance of silicon Raman devices. A desired reduction in lifetime is attained by reducing the lateral dimensions of the opti- cal waveguide, and by actively removing the carriers with a reverse biased diode. An integrated diode also offers the ability to electri- cally modulate the optical gain, a unique property not available in fiber Raman devices. Germanium-silicon alloys and superlattices offer the possibility of engineering the otherwise rigid spectrum of Raman in silicon. Index Terms—Nonlinear optics, Raman amplification, Raman laser, silicon, wavelength conversion. I. INTRODUCTION T HE recent wide-scale interest in silicon photonics can be identified with two motives. First, being able to tap into silicon’s vast manufacturing base will reduce the cost of pho- tonic devices which, in turn, will accelerate penetration of optics into communication at shorter distances than today’s fiber op- tic networks. Additionally, the technology can solve important problems in today’s computing systems, as well as spawn new industries of its own. For example, as the trend to reducing device dimensions continues, a significant bottleneck has ap- peared at the electronics interconnect level, where a large gap exists between individual device speeds and the speed of in- terconnects that link them [1], [2]. Optical interconnects can potentially solve this important problem. In the 1990s, a large number of passive silicon devices were developed [3] with a few reaching commercialization. However, due to unfavorable physical properties, such as the lack of efficient optical transitions due to the indirect band structure and the near-absence of Pockel’s effect caused by symmetric crystal structure, creation of active devices proved to be much more difficult. Manuscript received August 26, 2005. B. Jalali, V. Raghunathan, and D. Dimitropoulos are with the Department of Electrical Engineering, University of California, Los Angeles, CA 90095-1594 USA (e-mail: jalali@ucla.edu; varun@ee.ucla.edu; ddmitr@ee.ucla.edu). ¨ O. Boyraz is with the Department of Electrical Engineering and Com- puter Science, University of California, Irvine, CA 92697-2625 USA (e-mail: oboyraz@uci.edu). Digital Object Identifier 10.1109/JSTQE.2006.872708 The prospects for active optical functionality in silicon have drastically improved since the adoption of the Raman effect as a mechanism for producing amplifiers, lasers, and wavelength converters. Last year was witness to the demonstration of the first silicon laser [4]. The rapid pace of progress is continuing, and the first quarter of 2005 has already seen the demonstration of direct electrical modulation of the Raman laser [5] and report of the first continuous-wave (CW) silicon Raman laser [6]. Raman scattering was proposed and demonstrated in 2002 as a mean to bypass these limitations, and to create optical am- plifiers and lasers in silicon [7]. The approach was motivated by the fact that the stimulated Raman gain coefficient in silicon is 10 3 –10 4 times larger than that in fiber. The modal area in a silicon waveguide is roughly 100 times smaller than in fiber, resulting in a proportional increase in optical intensity. The com- bination makes it possible to realize chip-scale Raman devices that normally require kilometers of fiber to operate. The ini- tial demonstration of spontaneous Raman emission from silicon waveguides in 2002 was followed by the demonstration of stim- ulated Raman scattering [8] and parametric Raman wavelength conversion [9], both in 2003. Other merits of the Raman effect include the fact that it occurs in pure silicon and hence does not require rare earth dopants (such as Erbium), and that the spectrum is widely tunable through the pump laser wavelength. II. RAMAN SCATTERING IN SILICON Classical electrodynamics provides a simple and intuitive macroscopic description of the Raman scattering process [10]. In the spontaneous scattering, thermal vibrations of a lattice at frequency ω v (15.6 THz in silicon) produce a sinusoidal modulation of the susceptibility. The incident pump field in- duces an electric polarization that is given by the product of the susceptibility and the incident field. The beating of the inci- dent field oscillation ω p with oscillation of the susceptibility ω v produces induced polarizations at the sum frequency ω p + ω v , and at the difference frequency ω p − ω v . The radiation pro- duced by these two polarization components is referred to as anti-Stokes and Stokes waves, respectively. Quantum statistics dictates that the ratio of Stokes power to anti-Stokes power is given by (1 + N )/N , where N = [exp(¯ hω v /kT ) − 1] −1 is the Bose occupancy factor, and has a value of ∼0.1 for silicon at room temperature. The same model can be extended to describe stimulated Raman scattering [10]. Here, one assumes that pump and Stokes fields are present, with a frequency difference equal to the atomic vibrational frequency. The latter can be due to spontaneous emission, or in the case of a Raman amplifier, 1077-260X/$20.00 © 2006 IEEE