Traditionally, inverse design technology aimed to optimize fixed-shape optical structures which were based on basic shapes such as triangles or circles. However, the increasing demand for multifunctional and high-performance metasurface has brought about a need for a freeform design method that can handle a large design space with a magnitude of several orders higher than traditional structures. Here, we formulate the design problem of a beam deflector made of one-dimensional freeform silicon metasurface as a Reinforcement Learning (RL) problem. By utilizing RL, without a need for any prior metasurface data, we show that the suggested algorithm can derive optimal structures of the given problem with a simple neural network structure. During training, an agent designed as a deep Q-network randomly explores the design space and exploits its knowledge to optimize the deflection efficiency. The design strategy showed overall improvements in maximum efficiency when compared to state-of-the-art baseline approaches. Also, the algorithm proved its robustness by showing a small variance for multiple experimental initializations.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.