Increasing demands on data transfer have meant that the limits of electrical interconnects are being stretched to the point where an optical-based solution must take over. The loss on electrical interconnects increases with frequency. Therefore, as data rates are increased to cope with growing bandwidth requirements, the signal becomes more and more distorted. The capability of optical interconnects is far superior, with the possibility of transmitting multiple channels in excess of 10Gbit/s down a single fiber. Silicon offers low fabrication costs, the prospect of electronic-photonic integration, and compatibility with CMOS fabrication processes. It is therefore a strong candidate as a material within which to form circuits that integrate photonic components such as modulators, detectors, filters, and couplers to perform a certain function.
The optical modulator, which converts electrical data into an optical format, is the key element of such an optical interconnect. Traditionally, optical modulators have been made from materials that exhibit strong electro-optic effects. The lack of this attribute in silicon meant that, for a long time, the prospects of forming silicon-based high-speed optical modulators seemed dim. However, silicon's other advantages motivated significant research effort worldwide.
Modulation in silicon can be achieved in more than one way. For example, optical microelectromechanical systems (MEMS) switch light due to external physical movement, and thermo-optic devices use changes in temperature to alter the material's refractive index. These methods are too slow for high-speed optical modulation, however. A hybrid approach, where other materials such as III-V-type semiconductors (made by combining elements from groups III and V of the periodic table) are bonded, deposited, or grown on silicon is also possible. In this case, the light-guiding and other photonic functions are performed in the silicon layer, but in the modulation regions the light is coupled into the III-V material and its electro-optic effect used to modulate the light. However, hybrid approaches compromise CMOS compatibility and increase fabrication complexity. Plasma-dispersion-effect devices, in which electron and hole densities are electrically varied to cause changes in the complex refractive index, are therefore the most attractive.
Figure 1. Cross section of the rib (center) and slab regions (right and left) of the silicon-waveguide-based phase modulator.
There have been several noticeable breakthroughs in plasma-dispersion-effect devices. In 1986, Soref and Bennet demonstrated the first plasma-dispersion-effect optical modulator based on silicon.1 In 2004, Png et al. proposed the first carrier injection gigahertz modulator through simulations.2 The first experimental demonstration of a gigahertz modulator, which operated via carrier accumulation, was by Liu et al. the same year.3 In 2005, Gardes et al. proposed the first device based on carrier depletion,4 and in 2007, Liao et al. presented the first 40Gbit/s carrier depletion silicon optical modulator and demonstrated a 1dB extinction ratio. The extinction ratio describes the difference in power levels between the ‘1’ and ‘0’ data levels. A larger extinction ratio means that it is easier for the detector to resolve between the two levels, resulting in fewer errors.
Figure 2. The self-aligned fabrication process. (A) Boron is implanted into the active region to make it p-type. (B) A silicon dioxide (SiO2) layer is deposited and patterned with the waveguide design. (C) The SiO2 layer is used as a hard mask through which to etch the optical waveguides into the silicon overlayer. (D) The SiO2layer is used with a photoresist window with edge aligned anywhere on the waveguide rib as a phosphorus implant mask.
We have been working to push the performance to higher levels still. Within the European Union Seventh Framework project HELIOS, which started in 2008, we have been developing a novel high-performance modulator that will be simple to fabricate and show fewer performance variations.
Figure 3. Optical eye diagram obtained at 40Gbit/s showing an extinction ratio of 10dB.
Figure 4. Optical eye diagram obtained at 50Gbit/s showing an extinction ratio of 3dB.
The silicon waveguide rib (400nm width and 220nm height) and the slab region (100nm height) on one side are doped p-type (see Figure 1). The slab region on the other side is doped n-type. These n- and p-type regions extend out to p+- and n+-type regions, respectively, that form ohmic contacts to the coplanar waveguide electrodes that drive the device. Under reverse bias conditions, the depletion region at the junction of the p- and n-type regions becomes wider. By selecting appropriate doping concentrations, it is possible for the depletion to extend mainly into the rib region, where the majority of the optical power is carried.
Previously, several modulators based on this technology have had the p-n junction positioned in the center of the rib waveguide to maximize VπLπ efficiency (the voltage-length product required to achieve a π radian phase shift). However, the performance of the device is then sensitive to the position of this junction with respect to the waveguide, and alignment variations during fabrication will cause performance variations. A key feature of our device is that a self-aligned process can be used to form the p-n junction (see Figure 2). This avoids any such performance variations.
Many applications require intensity modulation rather than phase modulation. For this reason, we use Mach-Zehnder interferometers (MZIs) to convert between the two. We have measured passive extinction ratios of around 30dB from the fabricated MZIs.6 In DC, we have calculated the phase efficiency of the modulators as 2.7V·cm, and extinction ratios in excess of 25dB are achievable with a 6V drive on a 3.5mm-long device. We tested the high-speed performance of the device by observing its ability to convert electrical data streams at different data rates into an optical format. We have demonstrated optical modulation at 40Gbit/s with a record 10dB extinction ratio from a 3.5mm device (see Figure 3).6
We were also able to achieve modulation at 50Gbit/s (see Figure 4) from a device just 1mm in length.7 A modulation depth of 3dB was calculated when accounting for EDFA (erbium-doped fiber amplifier) noise. This is the first demonstration of 50Gbit/s modulation from a non-hybrid silicon optical modulator in the world. Such increases in data rate allow, first, for the bandwidth of a channel (one wavelength on one optical fiber) to be enhanced. Second, the power consumption per bit of data is reduced. Finally, if multiple channels are used to increase the aggregate data, either by sending more than one wavelength along one fiber (wavelength division multiplexing) or by using multiple fibers, the number of channels required is reduced.
In addition, we have obtained 40Gbit/s modulation from a slow-wave version of the device in collaboration with project partners at the Universitat Politècnica de València, Spain.8 The slow wave approach involves reducing the group velocity of the propagating light by using a 1D periodic structure. By slowing the light down we were able to significantly enhance the modulation effect and therefore enable a reduction in device length while enhancing the performance of the device.
In summary, silicon photonics could meet demand for high-speed optical interconnects. We have demonstrated silicon optical modulators that operate at data rates of up to 50Gbit/s. Furthermore, we have developed a fabrication process that reduces performance variations. We are now working to integrate the modulator with other photonic and electronic components.
The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/2007–2013) under grant agreement 224312 HELIOS and from the Engineering and Physical Sciences Research Council (UK) to support the UK Silicon Photonics project.
University of Southampton
Since 2008, David Thomson has been leading the work package on optical modulators within the European silicon photonics project named HELIOS. In 2012, he moved from the University of Surrey (UK) to the University of Southampton to further his research into silicon photonic devices.
Frederic Gardes, Youfang Hu, Graham Reed
School of Electronics and Computer Science
University of Southampton
Electronics and Information Technology Laboratory (LETI) Atomic Energy and Alternative Energies Commission (CEA)
1. R. A. Soref, B. R. Bennett, Kramers-Kronig analysis of electro-optical switching in silicon, Proc. SPIE 704, p. 32-37, 1986.
2. C. E. Png, S. P. Chan, S. T. Lim, G. T. Reed, Optical phase modulators for MHz and GHz modulation in silicon-on-insulator (SOI), J. Lightwave Technol. 22, p. 1573-1582, 2004.
3. A. Liu, R. Jones, L. Liao, D. Samara-Rubio, D. Rubin, O. Cohen, R. Nicolaescu, M. Paniccia, A high-speed silicon optical modulator based on a metal-oxide-semiconductor capacitor, Nature 427, p. 615-618, 2004.
4. F. Y. Gardes, G. T. Reed, N. G. Emerson, C. E. Png, A sub-micron depletion-type photonic modulator in silicon on insulator, Opt. Express 13, p. 8845-8854, 2005.
5. L. Liao, A. Liu, D. Rubin, J. Basak, Y. Chetrit, H. Nguyen, R. Cohen, N. Izhaky, M. Paniccia, 40 Gbit/s silicon optical modulator for high speed applications, Electron. Lett. 43, p. 1196-1197, 2007.
6. D. J. Thomson, F. Y. Gardes, Y. Hu, G. Mashanovich, M. Fournier, P. Grosse, J.-M. Fedeli, G. T. Reed, High contrast 40Gbit/s optical modulation in silicon, Opt. Express 19, p. 11507-11516, 2011.
7. D. J. Thomson, F. Y. Gardes, J.-M. Fedeli, S. Zlatanovic, Y. Hu, B. P. P. Kuo, E. Myslivets, N. Alic, S. Radic, G. Z. Mashanovich, G. T Reed, 50-Gb/s silicon optical modulator, Photon. Technol. Lett. 24, p. 234-236, 2011.
8. A. Brimont, D. J. Thomson, P. Sanchis, J. Herrera, F. Y. Gardes, J. M. Fedeli, G. T. Reed, J. Martí, High speed silicon electro-optical modulators enhanced via slow light propagation, Opt. Express 19, p. 20876-20885, 2011.