Bonn 2025 – wissenschaftliches Programm

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

Q: Fachverband Quantenoptik und Photonik

Q 70: Nanophotonics I

Q 70.7: Vortrag

Freitag, 14. März 2025, 12:30–12:45, HS I PI

Active Physics-Informed Deep Learning: Surrogate Modeling for Non-Planar Wavefront Excitation of Topological Nanophotonic Devices — •Fatemeh Davoodi — Institute for Experimental and Applied Physics, Kiel, Germany

Topological plasmonics provides innovative ways to manipulate light by combining principles of topology and plasmonics, akin to topological edge states in photonics. However, designing such states is challenging due to the complexity of the high-dimensional design space. In this work, we introduce a supervised, physics-informed deep learning framework combined with surrogate modeling to design topological devices for specific wavelengths. By embedding physical constraints into the neural network training process, our model efficiently navigates the design space, significantly reducing simulation time and computational cost.

Additionally, we incorporate non-planar wavefront excitations to probe topologically protected plasmonic modes, introducing nonlinearity into the design and training process. Using this approach, we successfully design a topological device featuring unidirectional edge modes in a ring resonator operating at specific frequencies. This method demonstrates the effectiveness of integrating machine learning with advanced physical modeling for photonic device innovation, achieving high accuracy while optimizing computational efficiency.

Keywords: Su-Schrieffer-Heeger (SSH) model; Topological Plasmonic; Non-Planar Wavefront Excitations; Physics-Informed Machine Learning; Deep learning