Molecular Nanocomputing: An Engineering Approach from Physics To Circuit Architectures

In the last 50 years, Moore's law has fueled the development of the computing field. Nowadays, skyrocketing fabrication costs and increasing demand for miniaturization, high performance and low power consumption are becoming incredibly challenging, pushing the research effort towards Beyond-CMOS technologies and innovative architectural solutions. In this scenario, molecular electronics paves the way for smart and cheap technological processes based on self-assembly. The possibility to produce artificial molecules with nano-scaled dimensions and engineered properties, is really attractive, since enables extremely high integration with fewer power issues accomplishing a variety of electronic tasks, including conducting wire, rectification, memory, sensor and switching only by simply changing the type of molecule. This thesis work investigates molecular nanocomputing solutions based on electronic transport, spacing from physical level to the application abstraction level, with constant attention devoted to providing a physical insight from an Electronics Engineer standpoint. The first part of this work deepens at the very physical level of abstraction in quantum transport models and the theory of conduction at nanoscale. With the aim to provide a practical understanding, theory's outcomes are aided in parallel by atomistic simulations. The second part is devoted to the device level of abstraction: different molecules from the state-of-art literature synthesized ad-hoc for computing applications are engineered in a device configuration and characterized by means of ab-initio simulations. Proper figures of merit are extracted. Finally, the aim of the last part is to investigate at architectural level how to overcome the limitations of present computational systems by exploiting the possibilities offered by molecular technology and unconventional architectures. Firstly a molecular implementation of logic (logic gates, Half- Adder, Full-Adder) and memory (SRAM) elements is performed, and then a target Logic-in-Memory application is discussed and implemented. Design, functional verifications, performance analysis are performed in Cadence Virtuoso both for molecular ambipolar transistors and FD-SOI MOSFET (28 nm) technologies in order to be compared and demonstrate the benefits and problems, for this specific application, in migrating from conventional to a molecular technology. At last, the stated goal of the thesis is achieved. It reaches an adequate awareness of how the nature of molecular technology is often very different from CMOS established one. As a consequence conventional architectural paradigms cannot really do justice of its overwhelming potentialities. Therefore possible alternative solutions are proposed and briefly analyzed in view of seeking out a deeper understanding of the physics of transport and storage inside molecules and "boldly go where no silicon-based computing system has gone before!"