Active optical devices in silicon-on-insulator rib waveguides.

Much progress has been made in the development of active silicon opto-electronic devices over the last 15 years. This is primarily due to the widely accepted belief that the free carrier effect is the most efficient optical modulation and switching mechanism in silicon, along with the potential advantages of combining optical and electronic devices onto a single silicon substrate rather than using discrete components. A study of the scientific literature shows that whilst numerous devices have been reported, few have been seriously optimised. In the literature, devices have consisted primarily of two or three terminal devices based around a rib waveguide. The three terminal devices are fewer in number but generally perform better. Conversely, two terminal devices have received a little more attention in terms of producing faster devices. Therefore, this work provides an in depth analysis of the performance of p+-i-n+ diodes when configured as optical modulators, with the aim of improving both the device DC and transient performance characteristics. The primary DC performance characteristic is the current required to achieve a given phase change and the transient performance characteristics are measured in terms of the device rise and fall times. These characteristics have been studied with variations in geometrical and fabrication based parameters such as the position and doping concentration of the contacts, the aspect ratio of the rib waveguides, and the overall dimensions. The key result from the modelling is that the most efficient multi-micron size device is a three terminal device with high doping concentration, constant doping profiles and large diffusion depth doped regions located close to the rib edge. A theoretical device of this nature required a current of only 2.7mA for a ? radian phase shift with rise and fall times of 22ns and 2ns respectively. The best previously achieved was a device which theoretically required 4mA for a ? radian phase shift. Additionally, by including isolation trenches on either side of the doped regions the DC performance characteristics can be further improved by up to 74%. There are also advantages in reducing the dimensions of the devices to 1 micron or less. At these dimensions the DC and transient performance characteristics are improved by more than a further order of magnitude, hence requiring fractions of 1mA for a ? radian phase shift. Two of the most promising designs have been fabricated and experimentally analysed. Due to fabrication constraints the most efficient device was not fabricated. However, both two and three terminal devices were fabricated. The best device tested experimentally was a three terminal device that required a current of 14mA for a ? phase shift. The modelling and experimental results agree well therefore validating the modelling. Therefore we can be confident that the additional theoretical results for devices that could not be fabricated are reliable, and hence significant further improvements could be made by fabricating these devices. Likely roles for these types of devices are medium bandwidth modulators/switches and a variety of sensor applications