Chromophore-based nanophotonic network-on-chip and computing systems

Traditional CMOS faces many scaling challenges as feature sizes approach tens of nanometers. Despite advances in next-generation transistor designs and foundry processes, power efficiency and thermal budgets continue to dominate the performance envelope of current systems. This motivates researchers to look into alternative technologies.;Molecular-scale computational devices such as chromophores are very promising. Chromophores are small molecules that absorb light at one wavelength and emit at a different wavelength with lower energy. They have such properties as low energy consumption and short latency. A pair of chromophores can use Resonance Energy Transfer (RET) to communicate with each other in near field. Simple RET logic circuits have been fabricated by patterning chromophores on DNA self-assembled grids.;We explore the design space of using chromophores to perform optical computing in an systematic way: 1) explore the optical filtering function of single types of chromophores in power efficient photonic crossbars; 2) discuss how to further reduce power consumption of the crossbars by utilizing the underlying hardware properties and application communication patterns; 3) design more complex general purpose computing systems with resonance energy transfer between different types of chromophores on DNA grids. We make three primary contributions in this thesis. First, we use only the filter function of chromophores to design an on-chip interconnection network---molecular-scale Network-on-Chip (mNoC)---and evaluate its performance and power consumption. Second, we explore the power property of mNoC and optimize the NoC by introducing the concept of power topology--- unique characteristics where connectivity is dynamically controlled by the source power. Finally, we propose a chromophore-based optical logic element (OLE) as the basic computing unit for a larger general computing system.