Advancing Silicon Photonics Through Machine Learning: From Device Design to Fabrication

The advances in fabrication technologies have enabled the development of new nanophotonic devices, utilizing improved structural resolution for greater integration density and functionality. The complexity of these devices necessitates the adoption of algorithm-assisted design methods to navigate the vast design space. This evolution presents a push-pull dynamic: while striving for advanced device functionalities, the limits of fabrication technology are tested, potentially affecting device robustness. Our research leverages machine learning to address this challenge, balancing the drive for innovation with the practicalities of manufacturing. This paper outlines our strategies to optimize device design and predict fabrication outcomes, improving both the performance and robustness of next-generation silicon photonic devices.