Abstract:The research of metamaterials has achieved enormous success in the manipulation of light in an artificially prescribed manner using delicately designed sub-wavelength structures, so-called meta-atoms. Even though modern numerical methods allow to accurately calculate the optical response of complex structures, the inverse design of metamaterials is still a challenging task due to the non-intuitive and non-unique relationship between physical structures and optical responses. To better unveil this implicit relationship and thus facilitate metamaterial design, we propose to represent metamaterials and model the inverse design problem in a probabilistically generative manner. By employing an encoder-decoder configuration, our deep generative model compresses the meta-atom design and optical response into a latent space, where similar designs and similar optical responses are automatically clustered together. Therefore, by sampling in the latent space, the stochastic latent variables function as codes, from which the candidate designs are generated upon given requirements in a decoding process. With the effective latent representation of metamaterials, we can elegantly model the complex structure-performance relationship in an interpretable way, and solve the one-to-many mapping issue that is intractable in a deterministic model. Moreover, to alleviate the burden of numerical calculation in data collection, we develop a semi-supervised learning strategy that allows our model to utilize unlabeled data in addition to labeled data during training, simultaneously optimizing the generative inverse design and deterministic forward prediction in an end-to-end manner. On a data-driven basis, the proposed model can serve as a comprehensive and efficient tool that accelerates the design, characterization and even new discovery in the research domain of metamaterials and photonics in general.

Automated Metamaterial Design with Computer Model Emulation and Bayesian Optimization

Data-Driven Design of Thermal-Mechanical Multifunctional Metamaterials

A Machine Learning-Based Method to Design Modular Metamaterials

Leveraging Parallel Speedup for Intelligent Design of Metasurface

RAPID DESIGN OF METAMATERIALS VIA MULTITARGET BAYESIAN OPTIMIZATION

An Efficient Robust Optimization Methodology for Metamaterial Unit Design

Machine Learning-Encoded Multiscale Modelling and Bayesian Optimization Framework to Design Programmable Metamaterials

Hierarchical Combinatorial Design And Optimization Of Quasi-Periodic Metamaterial Structures

Design of multifunctional metamaterials using optimization

Probabilistic representation and inverse design of metamaterials based on a deep generative model with semi-supervised learning strategy

Design Optimization of Metamaterial Units Using a Genetic Algorithm Based Optimization Methodology

An Improved Multi-Objective GA for Low Frequency Metamaterial Unit Robust Optimization under Uncertainty

Computational design of mechanical metamaterials

An intelligent design system for tailored metamaterial properties

Physics-Driven Machine-Learning Approach Incorporating Temporal Coupled Mode Theory for Intelligent Design of Metasurfaces

A Substructure Perturbation Method for Systematic Design of Mechanical Metamaterials with Programmed Functionalities

A Generative Model for Inverse Design of Metamaterials

Deep generative modeling for mechanistic-based learning and design of metamaterial systems

Inverse Design of Mechanical Metamaterials Using a Point-Cloud-Based Deep Generative Model

Robust Optimal Design of Low Frequency Metamaterials Considering the Fabrication Imperfections and Parameter Fluctuations

Optimal Design of One-Dimensional Elastic Metamaterials Through Deep Convolutional Neural Network and Genetic Algorithm