Light coupling between integrated photonic devices and the off-chip environment has always posed challenging research problems, especially for high-index-contrast plat forms such as Silicon-On-Insulator (SOI). Surface gratings can be flexibly arranged in any desired pattern on the chip and have been widely used to interface integrated circuits with both optical fibers and for free-space coupling applications. A number of solutions have been proposed to improve their efficiency in diffracting light but other limitations still exist. Dimensions of SOI surface gratings are typically on the order of few tens of microns and a significant size reduction is not easy to achieve without compromising efficiency. This can be a limiting factor for applications requiring a high integration density, e.g. optical phased arrays or multi-device photonic circuits with numerous fiber interfaces. Moreover, the diffraction angle of surface gratings normally depends rather strongly on the wavelength of the light, limiting their operational bandwidth compared to edge fiber couplers or end-fire antennas. Our research focuses in particular on the design of photonic antennas based on surface gratings very high efficiency, reduced size, and ultra-wide bandwidth. To achieve these goals we exploit complex unit cell geometries and subwavelength metamaterials and we make use of inverse design and machine learning techniques to optimize the structures.
D. Melati, Y. Grinberg, M. Kamandar Dezfouli, S. Janz, P. Cheben, J. H. Schmid, A. Sánchez-Postigo, Dan-Xia Xu, “Mapping the global design space of nanophotonic components using machine learning pattern recognition,” Nat Commun 10(1), 1–9 (2019) [http://doi.org/10.1038/s41467-019-12698-1].