Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 5: Poster

AKPIK 5.2: Poster

Donnerstag, 20. März 2025, 15:00–16:30, P2

Machine Learning Optimization of Chiral Photonic Nanostructures — •Davide Filippozzi¹, Nicolas Roy², Alexandre Mayer², and Arash Rahimi-Iman¹ — ¹I. Physikalisches Institut and Center for Materials Research, Justus-Liebig-Universität Gießen, 35392 Gießen, Germany — ²NaXys, Namur Institute for Complex Systems, University of Namur, Belgium

Deep learning (DL) and evolutionary algorithms (EA) as part of the machine learning (ML) domain have recently been well utilized for optimization purposes, such as for nanostructure design. Particularly, unintuitive problems can benefit from the potential abstraction levels that artificial Neural Networks (NNs) can achieve based on sufficient training and proper data. Reinforcement learning approaches promise to boost inference of solutions for complicated design requirements and specific functionalities.

We present a study that discusses the nano-pattern design optimization with a combination of DL and EA for a dielectric surface's preference for single-handed circularly polarized light in reflection or transmission. Advancing our previous simulations and algorithms [O. Mey and A. Rahimi-Iman, Phys. Status Solidi RRL 2022, 16, 2100571], the optimization in chiral dichroism and reflectivity for our metasurface's design is discussed. Such ML optimization can improve desirable features of unintuitive metamaterials and photonic nanostructures, as increasingly highlighted in up-to-date literature.

Keywords: Machine Learning; Artificial Neural Network; Evolutionary Algorithm; Design Optimization; Nanophotonics