High-performance computing and topology optimization for the design of large-scale dispersive nanophotonic structures in 3D.

• Applicant: Professor Dr. Antonio Calà Lesina
• Subject Area: Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale ; Optics, Quantum Optics and Physics of Atoms, Molecules and Plasmas ; Computer Architecture, Embedded and Massively Parallel Systems ; Theoretical Condensed Matter Physics
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 527470210

Nanophotonics studies the interaction of light with objects of nanoscale dimension, so-called nanostructures. Nanostructures arranged in a 3D or 2D lattice form a metamaterial or metasurface, respectively. Due to plasmonic resonances in metal nanostructures and Mie-type resonances in dielectric nanostructures, it is possible by design to achieve control of light at the nanoscale. Metasurfaces and metamaterials provide a revolutionary platform for flat optical devices, beam structuring, colouring, biosensing, nanomedicine, quantum communications, beam steering, nonlinear optics, and computing. Performing free-form optimization by means of high-performance computing has revolutionized many engineering disciplines, where new conceptual designs were obtained. However, its adoption in nanophotonics in combination with dispersive optical materials is missing. By exploring large design spaces one can find solutions to complex optimization problems, but also discover new conceptual designs. However, exploring such large parameter spaces is computationally challenging. Topology optimization (TopOpt) is a robust inverse design technique which produces designs far beyond human intuition and other optimization methods. In the optical regime, it is used by other groups to inverse design nanophotonic structures in the frequency-domain. My group has recently proposed a time-domain TopOpt algorithm for the inverse design of plasmonic nanostructures based on the Drude model. Time-domain TopOpt is computationally more efficient when dealing with wideband performance. Furthermore, time-domain methods are preferred for parallel computing due to their superior scalability. In this regard, a 3D-FDTD parallel code is available in my group and developed for over ten years. The first objective of this project is to equip our parallel 3D-FDTD code with parallel TopOpt functionalities to enable the inverse design of large-scale and free-form nanophotonic systems in 3D. The second objective is to develop a time-domain TopOpt algorithm for arbitrary optical dispersion to allow the design of dielectrics, semiconductors and metals in their interband transition region. Such techniques will then be employed in the third objective to design novel metasurfaces for efficient quantum light sources and ion-trapped quantum computing. Strategies based on deep learning will also be investigated in the fourth objective to mitigate the energetic impact of our computations.

• DFG Programme: Research Grants
• International Connection: Sweden
• Cooperation Partner: Professor Dr. Emadeldeen Hassan