A Universal Approach to Nanophotonic Inverse Design through Reinforcement Learning
Butz M; Leifhelm A; Becker M; Risse B; Schuck C
Research article in edited proceedings (conference)Abstract
We present a novel method to perform universal black-box optimization of pixel-discrete nanophotonic devices based on reinforcement learning. We demonstrate the capabilities of our method for a silicon-on-insulator waveguide-mode converter with > 95\% conversion efficiency.
Details about the publication
Editors: Optica Publishing Group
Book title: CLEO 2023, paper STh4G.3
Page range: STh4G.3-STh4G.3
Publisher: Optica
Place of publication: San Jose
Status: Published
Release year: 2023
Language in which the publication is written: English
Conference: CLEO: Science and Innovations 2023, San Jose, United States
Link to the full text: https://opg.optica.org/abstract.cfm?uri=CLEO_SI-2023-STh4G.3
Keywords: Effective refractive index, Evanescent wave coupling, Inverse problems, Mode conversion, Neural networks, Stochastic processes
Authors from the University of Münster
| Becker, Marlon Marijn | |
| Butz, Marco | |
| Risse, Benjamin | |
| Schuck, Carsten |