Multi-task topology optimization of photonic devices in low-dimensional Fourier domain via deep learning

Abstract

Silicon photonics enables compact integrated photonic devices with versatile functionalities and mass manufacturing capability. However, the optimization of high-performance free-form optical devices is still challenging due to the complex light-matter interaction involved that requires time-consuming electromagnetic simulations. This problem becomes even more prominent when multiple devices are required, typically requiring separate iterative optimizations. To facilitate multi-task inverse design, we propose a topology optimization method based on deep neural network (DNN) in low-dimensional Fourier domain. The DNN takes target optical responses as inputs and predicts low-frequency Fourier components, which are then utilized to reconstruct device geometries. Removing high-frequency components for reduced design degree-of-freedom (DOF) helps control minimal features and speed up training. For demonstration, the proposed method is utilized for wavelength filter design. The trained DNN can design multiple filters instantly and concurrently with high accuracy. Totally different targets can also be further optimized through transfer learning on existing network with greatly reduced optimization rounds. Our approach can be also adapted to other free-form photonic devices, including a waveguide-coupled single-photon source that we demonstrate to prove generalizability. Such DNN-assisted topology optimization significantly reduces the time and resources required for multi-task optimization, enabling large-scale photonic device design in various applications.

Deep Learning and Adjoint Method Accelerated Inverse Design in Photonics: A Review

“…To facilitate multi-tasks inverse design, a topology optimization method based on the DNN in the low-dimensional Fourier domain was proposed in [109]. The DNN took target optical responses as inputs and predicted low-frequency Fourier components, which were then utilized to reconstruct device geometries.…”

On‐Chip Electrically Driven Tunable Meta‐Lens for Dynamic Focusing and Beam Steering

“…[1][2][3][4] A wide variety of photonic structures have been demonstrated to interact with and manipulate the propagating DOI: 10.1002/lpor.202300330 waves in planar waveguides. [5][6][7][8] Most of these functionalities rely on guiding wave behaviors in single mode or few mode waveguides with lateral confinement. There are also some structures working far from single mode condition like on-chip lens, [9,10] mirrors, [11,12] and curve gratings, [13,14] which enable unique capabilities of beam shaping and parallel signal processing.…”

Efficient Waveguide Taper Design with Rapid and Data-Driven Eigenmode Expansion

“…In recent years, machine learning-based design methods have become popular for designing compact and highperformance photonic devices [9]. In these approaches, a design space consisting of geometrical parameters of the device is sampled and searched for the most suitable optical response [10,11]. These approaches are typically coupled with FDFD [12] or FDTD [13] simulations that estimate individual devices' performance during this optimization process.…”