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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 5: Poster

AKPIK 5.10: Poster

Donnerstag, 12. März 2026, 15:00–16:30, P5

Advancing Machine Learning Optimization of Chiral Photonic Metasurface: Comparative Study of Neural Network and Genetic Algorithm Approaches — •Davide Filippozzi¹, Alexandre Mayer², Nicolas Roy², Wei Fang³, and Arash Rahimi-Iman¹ — ¹I. Physikalisches Institut and Center for Materials Research, Justus-Liebig-University, Gießen, Germany — ²Department of Physics, Namur Institute for Complex Systems (naXys), University of Namur, Belgium — ³College of Optical Science and Engineering, Zhejiang University, Hangzhou, China

We report on an advanced optimization framework for chiral photonic metasurfaces, comparing a refined Neural Network (NN) pipeline against a Genetic Algorithm (GA). By introducing a two-output NN architecture and exploiting geometric symmetries for data augmentation, we successfully reduce the trade-off between circular dichroism (CD) and reflectivity. Our comparative analysis on GaP and PMMA structures reveals complementary strengths: the GA excels in finding global optima for complex geometries, while the NN provides superior computational efficiency for large-scale screening. The optimized designs demonstrate a close to twofold increase in CD compared to Ref. [Mey & Rahimi-Iman, PSS-RRL 16, 2100571 (2022)]. We propose a hybrid workflow combining both methods to accelerate the design of effective chiral mirrors for polarization-selective light-matter interaction studies.

Keywords: Machine Learning; Neural Networks; Genetic Algorithm; Optimization; Metasurfaces