Monolithic and hybrid nanophotonic chips for high-speed and power-efficient optical computing and interconnects

Past decades have witnessed the unprecedented success in very-large scale integration-based electronic computing, as indicated by the Moore’s law. Transistors continue to scale down from micrometer to nanometer and computing cores continue to increase with all the benefits of lower power consumption and higher speed coming naturally until the falter of Moore’s law in recent years as nanofabrication technology approaches its inevitable limit. Remarkable efforts have been made to develop high-speed and power-efficient computing alternatives in this post-Moore’s law era. The exploration of optical computing started as early as ~1970s when integrated electronic computing was just about to take off. The continuous success of electronic computing over decades has not slowed down but instead stimulated the research on optical computing. It is because photonics own unique and fascinating properties such as ultrahigh speed, high bandwidth, and various multiplexing techniques, making it a promising candidate to supersede electronics especially when electronics face unsolvable obstacles. Fortunately, in this post-Moore’s law era, integrated photonics has evolved rapidly with abundant passive and active micrometer-size components available with ultralow power consumption. It enables integrated optical computing to be comparable with transistor-based electronic computing in terms of speed and power. In this dissertation, a new architecture of electronic-photonic computing will be presented, which makes the full use of electronics and photonics to achieve ultrahigh-speed and low-power-consumption computing that are comparable with the state-of-the-art transistors-based electronic computing. This dissertation will cover bottom-to-top discussions, ranging from gates, circuits, to automatic design algorithm, to architecture and entire system. First, it starts with the introduction of electro-optic logic and then the design and analysis of fundamental logic gates and circuits along with the essential components such as electro-optic modulators. Second, an automated logic synthesis algorithm called And-Inverter Graph will be presented, which is capable of designing computing circuits using these fundamental logic gates automatically and efficiently. Third, an electronic-photonic arithmetic logic unit is proposed with an experimental demonstration of 4-bit circuit operating at 20 GHz. Thorough discussion of the performance is also conducted which shows its potential to surpass the state-of-the-art transistors-based computing circuits in terms of speed and power. Forth, a Moore’s law in optical computing is discussed with one promising method called multi-operand logic gates provided. Lastly, architectural considerations of electronic-photonic computing are presented.Electrical and Computer Engineerin