Regensburg 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 4: Focus: Applications of Deep Neural Networks
AKPIK 4.3: Hauptvortrag
Dienstag, 18. März 2025, 15:00–15:30, H5
Inverse design of lateral hybrid metasurfaces with machine learning — •Rui Fang1, Amir Ghasemi1, Dagou Zeze1, Koen Valk2, Yuqing Jiao2, Peter Zijlstra2, and Mehdi Keshavarz Hedayati1 — 1Durham University — 2Eindhoven Technology University
The development of metasurface structural colour typically depends on laborious and time-consuming simulations such as Finite Element Method (FEM) or Finite-Difference Time-Domain (FDTD) simulation, along with human intuition for parameter adjustments, rendering it impractical for design optimization. In this context, we have introduced an innovative AI-assisted design process that circumvents the intricate simulations, allowing for a swift and precise correlation between metasurface parameters and colour coordinates. In this study, we have applied the model to the lateral hybrid design, a novel concept in metasurfaces proposed by our research group and demonstrated that the model can predict a structure tailored to achieve continuous colour coordinates with an accuracy of up to 97%. A noteworthy aspect of our discovery is that the model is capable of predicting the range of colours that can be generated from a single design of an active metasurface. Our deep learning approach proves to be a valuable tool in designing active metasurfaces for structural colours. This advancement contributes to the development of highly sensitive sensors, bringing tunable metamaterials closer to practical applications.
Keywords: metasurface; structural colour; machine learning; inverse design