Deep Learning Enabled Strategies for Modeling of Complex Aperiodic Plasmonic Metasurfaces of Arbitrary Size

Abstract

Optical interactions have an important impact on the optical response of nanostructures in complex environments. Accounting for interactions in large ensembles of structures requires computationally demanding numerical calculations. In particular, if no periodicity can be exploited, full field simulations can become prohibitively expensive. Here we propose a method for the numerical description of aperiodic assemblies of plasmonic nanostructures. Our approach is based on dressed polarizabilities, which are conventionally very expensive to calculate, a problem which we alleviate using a deep convolutional neural network as surrogate model. We demonstrate that the method offers high accuracy with errors in the order of a percent. In cases where the interactions are predominantly short-range (e.g., for out-of-plane illumination of planar metasurfaces), it can be used to describe aperiodic metasurfaces of basically unlimited size, containing many thousands of unordered plasmonic nanostructures. We furthermore show that the model is capable to spectrally resolve coupling effects. The approach is therefore of the highest interest for the field of metasurfaces. It provides significant advantages in applications like homogenization of large aperiodic planar metastructures or the design of sophisticated wavefronts at the micrometer scale, where optical interactions play a crucial role.

Generalizing the exact multipole expansion: Density of multipole modes in complex photonic nanostructures

“…Finally, we anticipate that the very low storage requirements will allow to use the generalized polarizabilities efficiently in lookup tables and also together with deep learning for various applications ranging from the interpretation of the optical properties of individual nanostructures to the design of complex metasurfaces. [53][54][55][56]…”

Incorporating Meta‐Atom Interactions in Rapid Optimization of Large‐Scale Disordered Metasurfaces Based on Deep Interactive Learning

“…To overcome the bottleneck, machine learning (ML) approaches have been introduced to CEM in recent years. [24][25][26][27][28][29][30][31][32][33] As one of the mainstream approaches nowadays in data classification and regression problems, the deep learning (DL) technique can efficiently and accurately solve EM inverse problems where scattering EM fields are used as the input and the ground-truth scatterer as the output. [24,26] Moreover, it can also be used to solve both forward and backward problems of metasurface design.…”

“…[ 17 ] Additionally, the deep neural network (DNN)‐based method of resolving optical coupling on aperiodic nanostructures of arbitrary size is also investigated. [ 30 ] Another novel DL architecture, which is widely used in the design of metasurfaces, is the generative adversarial network (GAN). [ 31 ] Coupled with the inverse modeling problem is metasurface optimization using traditional algorithms, such as genetic algorithm (GA) and particle swarm optimization (PSO), against spectral or radiation requirements.…”

Equivalent Circuit Models of a Bifunctional Optical Metasurface for Beam Splitting and Refractive Index Sensing

“…Compared with traditional three-dimensional metamaterials, the two-dimensional metasurface, composed of a single-layer meta-atom array, can not only avoid complex fabrication processes but also complete electromagnetic modulation within a subwavelength scale . The design, analysis, and optimization of metasurfaces are mostly based on the classical electromagnetic theory; thus, the full-wave numerical simulation technology can provide precise optical responses from metasurfaces with arbitrary meta-atom structures and arrangements through solving Maxwell’s equations. − However, the numerical results require substantial calculating time and take up much computer memory. − On the other hand, they cannot illustrate the physical mechanism behind the light–matter interactions …”