Evolution of optical modulation in silicon-on-insulator devices

One of the main goals of silicon photonics is to enable the fabrication and integration of electronic and photonic components on the same chip using the existing complementary metal -oxide -semiconductor (CMOS) processing platform. Due to its intrinsic structural properties, silicon does not exhibit a useful electro-optic effect that could enable light modulation. However, what makes silicon an attractive optical material is its transparency to infrared communication wavelengths and its high refractive index which facilitates the miniaturization of photonic devices. This also enables a high level of light confinement in nanometer-sized waveguides and provides an excellent basis to fabricate micro-optical devices. In silicon, light modulation is usually achieved via the free-carrier plasma dispersion effect, in which a change in carrier concentration (holes and electrons) is used to change the refractive index of the semiconductor, which, in turn, modifies the propagation velocity of light and the absorption coefficient. In resonators or interferometers, this effect enables the fabrication of modulators based on silicon-on-insulator (SOI) technology, which uses a layered silicon-insulator-silicon substrate in place of conventional silicon substrates.

Design rules for optical modulators in silicon should take into account single-mode waveguide design, large bandwidth operation, high modulation speed, high extinction ratio, small device size and low power consumption. Many different approaches have been used to satisfy these design goals on silicon platforms. Driven by these requirements, researchers have recently significantly reduced the waveguide core size from the micrometer range, producing silicon nanowires with a height and width of about 500nm and less. During the past four years, the scaling of waveguides using the plasma dispersion effect has enabled optical modulation frequencies to evolve in silicon from the MHz range to approximately 30GHz. This substantial leap in operating frequency now allows silicon-based optical devices to match the high-speed data transfer requirements of networking, board-to-board, or chip-to-chip interconnect applications.

Some of the most promising devices based on the plasma dispersion effect use injection of carriers (see Figure 1), accumulation of carriers on both sides of an insulating region (see Figure 2), and positive-negative (PN) junction depletion (see Figure 3) configurations.

Figure 1. Positive-intrinsic-negative (PIN) junction of an injection-type modulator.

Figure 2. Metal-oxide-semiconductor (MOS) capacitor of an accumulation-type modulator.

Figure 3. PN junction of a depletion-type modulator.

Injection modulators are based on the injection of carriers (holes and electrons) in the waveguide core. The modulation speed, however, is limited by the recombination lifetime and by the physical dimensions of the waveguide which requires the resistive contacts (highly doped regions) to be only a few hundred nanometers away from each other to avoid limiting the bandwidth. High-speed interferometers (see Figure 4) based on this concept with a bandwidth of 1GHz were first proposed in 2004,¹ followed in 2005 by a design featuring a much reduced waveguide size and significant bandwidth increase (24GHz), achieved by insertion of “lifetime killers” (impurities) in the waveguide area.²

Figure 4. SOI Mach Zehnder interferometer with an optical modulator inserted in one arm.

Metal-oxide-semiconductor (MOS) capacitor-type modulators are accumulation devices in which only majority carriers are accumulated on both sides of an insulating layer of silicon dioxide. This layer, placed in the waveguiding region, limits the bandwidth of the device to the resistance-capacitance cut-off frequency. An important milestone in silicon photonics was reached by Intel in 2004 with the first demonstration of a micrometer-size MOS modulator on SOI with a modulation bandwidth exceeding 1GHz.³ This breakthrough was quickly followed by experimental devices with a bandwidth improved to 6 GHz.⁴ Smaller MOS-type devices have also been proposed in 2006, with a six-fold increase in efficiency for an equivalent data rate.⁵

Depletion-type modulators are based on a PN junction located in the waveguide core. In these devices, the P-type region is designed to have a higher overlap with the optical mode than the N-type region (as the change in refractive index is higher for holes for a given density). The expected theoretical intrinsic bandwidth for this type of modulator is expected to exceed 50GHz. In 2005, a Mach Zehnder interferometer (MZI) modulator design based on the depletion of a PN junction inserted in a sub-micrometer waveguide was proposed with a theoretical bandwidth exceeding 50GHz.⁶ A micrometer-sized waveguide based on this principle was subsequently proposed, enabling polarization-independent performance of the modulator.⁷ In 2006, 10 Gb/s modulation based on a depletion device was also demonstrated in a MZI. In early 2007, the principle of operation of a depletion modulator was demonstrated in a submicron waveguide, achieving 20GHz-bandwidth modulation, recently improved to 30GHz.^8,9 A lateral junction depletion modulator with a bandwidth exceeding 1GHz has also been reported¹⁰ with some verbal reports mentioning 10GHz modulation.

Using the plasma dispersion effect, SOI optical modulation achieving 40Gb/s has now become reality. However, research still has to improve the power requirements at high frequency as well as the extinction ratio, currently of 1dB at 20GHz.⁸

Two other very promising effects were recently reported in silicon which could very well replace plasma dispersion. The first is a linear electro-optic effect which was first demonstrated in 2006.¹¹ It is induced by breaking the silicon crystal symmetry when a silicon nitride straining layer is deposited on the top of the waveguide. The second is the electro-absorption quantum confined Stark effect, which was first demonstrated in strained Ge/Si multiple quantum wells in 2005.¹² No high-speed measurements were presented, but modulations exceeding 50GHz were reported in other materials.¹³

It is clear that silicon optical modulators have undergone a significant transformation during the last four years. Device dimensions have been reduced to allow operation at the market standard and up to 40Gb/s,⁹ allowing silicon to become a choice candidate material for applications such as networking, chip-to-chip, and intra-chip communications. This is a significant achievement in a material that does not intrinsically exhibit a significant electric field-based modulation mechanism. While these devices still require improvement to displace conventional optical modulators, the possibility of producing all-silicon optical circuits and subsequently integrating them with electronic circuits represents a huge potential advantage for silicon. These recent advances not only increase the application areas of silicon but are also geared to enhance the performance of next-generation electronic circuits. Hence, silicon photonics is enabling silicon to become the photonic material of the future, and perhaps the platform upon which the next technological revolution will be built.