Researchers are getting better at making silicon do what it really does not like to do — emit light. A silicon laser is now demonstrated that has promising features for future practical applications.
Nonlinear optics could be called the optical equivalent of the philosopher's stone: just as lead could be turned into gold by changing the number of protons in its atoms, so the colour of a laser beam can be changed from blue to red by crossing a nonlinear crystal. In nonlinear optics, incident light is converted to light of a different wavelength by making use of specific, nonlinear properties of a material. Recently, nonlinear optics has been found to be capable of performing another much sought-after trick — transforming silicon, the main material for electronics, into an optically active material. On page 725 of this issue, Rong et al.1 take a further step towards a practical implementation of silicon optics by building a silicon laser that operates in a stable, continuous mode.
In the same way that steel is the base material for large-scale constructions and carbon for all known life forms, silicon is the mainstream material of electronics. It exhibits the right electronic and mechanical properties, is cheap and abundant, and can be easily processed into high-quality micrometre-scale devices. One thing that silicon could not do until recently was generate light efficiently, because of the nature of its electronic states. For this reason, all active optoelectronic applications, such as lasers for optical recording or for telecommunications, are based on group III–V materials such as GaAs and InP (ref. 2). However, as the clock frequencies of computer processors continue to increase, there is a growing need for optical data transmission that is integrated within silicon chips. Clock signals, which are needed to synchronize functions on a chip, are generated and trans- transmitted electronically, but this scheme will run into problems with power consumption and accuracy at higher processing speeds. Optical, rather than electronic, clock distribution is expected to circumvent these problems, and the achievement of practical optical amplification in silicon would therefore be a significant advance.
The nonlinear optical effect that is used to induce light emission and amplification (laser action) in silicon is ‘stimulated Raman scattering’ (Fig. 1a). A laser ‘pump’ activates the process; a photon at the pump energy is absorbed and then re-emitted with lower energy (and so a longer wavelength) together with a ‘phonon’ — an elementary vibration of the crystal3. The emitted photons make up the signal beam. By a trick of quantum wizardry, the upper energy level (dashed line in Fig. 1a) may remain virtual so that no real optical absorption is needed and the silicon crystal remains transparent. Laser action occurs because the process of light emission is stimulated — boosted — by the presence of a signal beam photon. The result is that the energy from the pump laser is transferred to the signal beam, which is then amplified.
‘Raman amplification’ is a small effect, and to build a laser with it you need a very high pump intensity and very low absorption losses. Such conditions have already been achieved in optical devices made from silica (SiO2)(refs 4, 5). To achieve a sufficiently large optical intensity to produce the Raman effect in silicon, Rong et al.1 used a recently developed silicon technology called silicon-on-insulator6. In this approach, which was originally designed to reduce the power consumption in portable electronics, thin layers of crystalline silicon with a large refractive index (n=3.6) are deposited on silica layers with a low refractive index (n=1.5). The large step in refractive index enables a tight confinement of light, which can be exploited to achieve significant Raman amplification in silicon.
Using this approach, Rong et al. built a silicon waveguide structure, in the shape of a ridge, surrounded by silica to guide the light with low losses (Fig. 2). They show that a pump laser with a power of only a fraction of a watt focused into this silicon waveguide creates an optical intensity up to 25 MW cm−2, which is larger than what has been achieved within high-power semiconductor lasers. The Raman amplification at this intensity is still small (a few decibels per centimetre, compared with 200 dB cm−1 in standard semiconductor lasers) but is enough to produce laser action, owing to low optical losses in the silicon waveguide.
Raman amplification has already been shown in similar silicon structures, but the amplification was limited to very short pulses of a few nanoseconds at most7,8. The problem is that an unwanted nonlinear optical side effect — two-photon absorption (Fig. 1b) — creates pairs of electrons and holes that remain for a long time in the sample and absorb both the pump light and the signal light, and so quickly turn off the Raman amplification. Rong et al. solve this problem by embedding the silicon waveguide within a semiconductor device, a reverse-biased p-i-n junction diode (Fig. 2). This device is designed to extract electrons and holes away from the waveguide. It is formed by implanting a short region on each side of the waveguide with impurities that convert silicon into a material with electron (n-side) or hole (p-side) conduction. A positive voltage is then applied to the n-side with respect to the p-side. In this reverse-biased scheme no current flows, but a strong electric field is generated that quickly removes the electrons and holes created by the two-photon absorption effect.
With this design Rong et al. demonstrate a silicon laser with continuous operation, a significant advance for the development of practical silicon lasers. Of course, the use of a nonlinear optics phenomenon means that optical pumping will always be required. However, the technique converts only a small fraction of the pump power to heat in the chip, in contrast to optically pumped lasers that do not rely on nonlinear optical effects. This is an important advantage given that heat dissipation is becoming the key limiting parameter in microelectronics.
A fascinating feature of this work is the use of the p-i-n junction, which combines the nonlinear-optical and semiconducting properties of silicon in the same device. Rong et al. show that this design enables control of the optical power emitted by the laser, which in principle should also be possible at a very high frequency and could therefore be used for information processing. Last but not least, this work demonstrates that technological advances in microelectronics, in this case the silicon-on-insulator and nanolithography techniques used to fabricate the waveguide ridge structure, can be applied to create advances in an apparently unrelated research field such as optoelectronics.
Rong, H. et al. Nature 433, 725–728 (2005).
Rosencher, E. & Vinter, B. Optoelectronics (Cambridge Univ. Press, 2002).
Shen, Y. R. The Principles of Nonlinear Optics (Wiley, New York, 1984).
Spillane, S. M., Kippenberg, T. J. & Vahala, K. J. Nature 415, 621–623 (2002).
Kippenberg, T. J., Spillane, S. M., Armani, D. K. & Vahala, K. J. Optics Lett. 29, 1224–1227 (2004).
Luryi, S., Xu, J. & Zaslavsky, A. (eds) Future Trends in Microelectronics (Wiley-IEEE, New York, 2004).
Boyraz, O. & Jalali, B. Optics Express 12, 5269–5273 (2004).
Rong, H. et al. Nature 433, 292–294 (2005).
About this article
AIP Advances (2019)
Theoretical Computer Science (2018)
Journal of Nanophotonics (2017)
Future Generation Computer Systems (2013)
The European Physical Journal Applied Physics (2012)