ABSTRACT
In this chapter, we introduce a physics-guided optimization approach to design photonic devices for high-performance computing systems. As motivation for one aspect of high-performance computing systems that may benefit from this approach, we begin by presenting a rough estimation of how improvements to the interconnect performance in a large computing system translate to improvements in the overall system performance. These results point to large system-level gains that could be accessed with optical interconnects, if better photonic components could be designed. In the rest of the chapter, we discuss the inverse-design method and its potential for designing photonic devices that are compact, highly efficient, and robust to fabrication and environmental factors, enabling the realization of practical chip-scale optical interconnects.
