Fast Filter Synthesis and Placement for Programmable Photonics

Abstract

Programmable photonics is a rapidly evolving field aimed at improving the accessibility and versatility of photonic integrated circuits. Despite recent advances, a major challenge persists: the lack of efficient and scalable algorithms for configuring the circuits to realize complex programmable functionalities. We propose a comprehensive methodology for routing and filtering in programmable photonic integrated circuits. Our approach begins with a high-level description of the filtering and routing requirements and systematically generates a fully functional configuration that satisfies all criteria. Central to our approach is an analytical formulation of the filter response, which not only facilitates rapid evaluation but also enables efficient derivative computation. We employ a divide-and-conquer strategy to calculate these derivatives with little additional overhead. This allows us to utilize derivative-based minimization techniques for filter synthesis, thereby reducing computation time. Furthermore, we demonstrate a simulated-annealing-based placement algorithm that focuses on concurrently placing multiple filters. Prior work has largely focused on small-scale demonstrations, typically involving one or two filters simultaneously. To the best of our knowledge, this is the first method capable of configuring photonic meshes that support the simultaneous realization of over 20 distinct light-processing functions, while maintaining practical run times. In this paper, ChatGPT and Claude were used for editing and grammar enhancement.