Designing nanophotonic structures using conditional deep convolutional generative adversarial networks

Abstract

Data-driven design approaches based on deep-learning have been introduced in nanophotonics to reduce time-consuming iterative simulations which have been a major challenge. Here, we report the first use of conditional deep convolutional generative adversarial networks to design nanophotonic antennae that are not constrained to a predefined shape. For given input reflection spectra, the network generates desirable designs in the form of images; this form allows suggestions of new structures that cannot be represented by structural parameters. Simulation results obtained from the generated designs agreed well with the input reflection spectrum. This method opens new avenues towards the development of nanophotonics by providing a fast and convenient approach to design complex nanophotonic structures that have desired optical properties.Keywords Nanophotonics · Inverse design · Conditional deep convolutional generative adversarial network · Deep learning Recently, data-driven design approaches have been proposed to overcome this problem. These approaches use artificial neural networks to design nanophotonic structures [10,11,12,13]. Previous studies first set the shape, such as multilayers [10] or H-antenna [11] of structures to be predicted, then trained NNs to provide output structural parameters that achieve desired optical properties. Once the NNs are trained, they provide the corresponding design parameters without additional iterative simulations. Such attempts have greatly reduced the effort and computational costs of designing nanophotonic structures. So far, these approaches have only been applied to conditions in which

Mid-Infrared Grayscale Metasurface Holograms

“…From the raw data of the measured image, we estimate that the unwanted zeroth-order perturbation is 19%, indicating that the total efficiency of our metasurface-based hologram is considerably high, up to 81%. We would like to underline the fact that the reported performance could be further increased, notably by properly considering near-field coupling of metasurface building blocks using optimization design methods [30][31][32][33][34][35].…”

Role of Artificial Intelligence and Machine Learning in Nanosafety

“…In related fields, Kim et al recently reported the use of GANs in the inverse design of porous materials [68] and So and Rho used deep convolutional GANs to generate new nanophotonic structures by inverse design. [69] Gomez-Bombarelli et al showed how a DNN and a recurrent neural network (RNN) can be used as an encoder and the decoder, respectively, for inverse design. The DNN models structure-property relationships between molecule structures and material properties and encodes the materials into latent molecule descriptors encoding molecule information.…”

Metasurfaces: Subwavelength nanostructure arrays for ultrathin flat optics and photonics

“…Recently, machine learning (ML) has emerged as an alternative to solve complex inverse design problems of metasurfaces, [46][47][48][49][50][51] as discussed in the Malkiel et al article in this issue. 52 In ML, which is a subset of the field of artificial intelligence, the system learns a mathematical complex model by consulting past experience or example data.…”

“…An alternative approach involving designing structural images instead of structural parameters has extended the degree of freedom significantly (Figure 2d). [50][51]56 Structural cross-sectional images are prepared as output forms; as a consequence, arbitrary shapes can be designed. The arbitrary shapes of such designed structures provide for the possibility that inverse design using ML could extend well beyond human intuition.…”