Inverse Design of Nanophotonic Devices using Dynamic Binarization

Abstract

The complexity of applications addressed with photonic integrated circuits is steadily rising and poses increasingly challenging demands on individual component functionality, performance and footprint. Inverse design methods have recently shown great promise to address these demands using fully automated design procedures that enable access to non-intuitive device layouts beyond conventional nanophotonic design concepts. Here we present a dynamic binarization method for the objective-first algorithm that lies at the core of the currently most successful inverse design algorithms. Our results demonstrate significant performance advantages over previous implementations of objective first algorithms, which we show for a fundamental TE00 to TE20 waveguide mode converter both in simulation and in experiments with fabricated devices.

Accelerating finite-difference frequency-domain simulations of inverse designed structures in nanophotonics using deep learning

“…We also expect a higher sample-efficiency as well as improved quality of the predictions when the model is trained on device-field combinations, additionally featuring continuous rather than just discrete permittivity distributions. These structures, although not representing realizable devices, are often encountered in established inverse design routines [9] and provide additional insights, for example related to wavelength-permittivity relations. Moreover, analytical inverse design methods often rely on the calculation of adjoint fields, which the model could also be trained on.…”

“…Being substantial components of photonic integrated circuits, the efficiency of these elements has great impact on the system's overall performance, for example in optical communication [25,26], optical phased arrays [27,28] or signal processing in complex chip layouts [29]. The optimization was conducted using an autonomously learning agent [30] where a splitting ratio of 90 by 10 for a wavelength of 𝜆 = 775 nm was cosen as the objective. We here consider the 100 nm tantalum-pentoxide-oninsulator platform, which has attractive properties for nonlinear and quantum photonics [2,31,32].…”

“…The range of applications spans from efficient fiber-tochip coupling [3][4][5] over complex multi-port sub-wavelength interference-based devices [6] to the realization of well established fundamental circuit components in novel and promising, yet challenging material platforms [7]. Current efforts address demands for higher device performance and more complex functionalities, requiring either larger device footprints [8] or smaller feature sizes [9] to increase the solution space of the optimization problem. The correspondingly larger simulation mesh imposes challenging requirements onto the electromagnetic simulation routines employed in the evaluation stages of iterative design processes.…”