[10a-B11-1~12] 3.10 Photonic structures and phenomena

10:15 AM - 10:30 AM JST(1:15 AM - 1:30 AM UTC)Presentation by Applicant for JSAP Young Scientists Presentation Award

[10a-B11-6] Photonic band calculation of photonic crystals using subspace learning

〇Takahiro Uemura¹, Kazuo Aoyama², Hisashi Sumikura^1,3, Masaya Notomi^1,3 (1.NTT BRL, 2.NTT CSL, 3.NTT NPC)

Keywords:

Photonic Crystal,Machine Learning,Physics-informed neural network

The photonic band structure, which characterizes the properties of photonic crystals, is generally calculated by solving eigenvalue problems obtained using finite-difference methods or basis expansions. Even in two-dimensional calculations, this requires repeated solutions of eigenvalue problems for large matrices consisting of several hundred or more basis functions, resulting in a high computational cost. Here, we report a method that learns a subspace of the eigenspace using a multilayer perceptron (MLP), a fundamental component of deep learning, and compresses the basis size by more than an order of magnitude.