Accelerating Design and Enabling High-Volume Adoption of Planar Lightwave Circuits Through Machine Learning

We introduce advancements in integrated photonic designs using a silica-on-silicon planar lightwave circuit platform. The platform exhibits ultra-low propagation losses, temperature-stable operation, and tight polarization control, allowing the realization of advanced architectures for optical communication, LiDAR and optical computing applications. Leveraging machine learning, we optimize design parameters on production masks to achieve performance uniformity. Furthermore, we use support vector machines to predict device performance on a wafer scale. These approaches accelerate photonic chip design and address challenges posed by fabrication variations, while fostering widespread adoption of photonic solutions in high-volume applications.