A Generative Model for Inverse Design of Metamaterials

By Zhaocheng Liu, Dayu Zhu, S. Rodrigues, Kyu-Tae Lee, Wenshan Cai. 2018

642 citations recorded by the source corpus.

The advent of metasurfaces in recent years has ushered in a revolutionary means to manipulate the behavior of light on the nanoscale. The design of such structures, to date, has relied on the expertise of an optical scientist to guide a progression of electromagnetic simulations that iteratively solve Maxwell's equations until a locally optimized solution can be attained. In this work, we identify a solution to circumvent this conventional design procedure by means of a deep learning architecture. When fed an input set of customer-defined optical spectra, the constructed generative network generates candidate patterns that match the on-demand spectra with high fidelity. This approach reveals an opportunity to expedite the discovery and design of metasurfaces for tailored optical responses in a systematic, inverse-design manner.

Subfields: Developmental Biology, Electronic, Optical and Magnetic Materials, Media Technology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Developmental Biology

1. Cynthia Rudin

In Electronic, Optical and Magnetic Materials

1. Yongmin Liu
2. Wei Ma
3. Yihao Xu
4. Rohit Unni
5. Zhicheng Wang
6. Chiara Daraio
7. Lu Lu
8. Kan Yao

In Media Technology

1. Zhi Chen
2. Yun Raymond Fu
3. Cynthia Rudin
4. Radu Timofte
5. Kai Zhang
6. Lei Zhang
7. Chao Dong
8. Luc Van Gool

Citations in the corpus

• How to See Hidden Patterns in Metamaterials with Interpretable Machine Learning (2021, methodology): “Model Frequency range [0, 10]kHz [10, 20]kHz [20, 30]kHz [30, 40]kHz [40, 50]kHz”
• Probabilistic Representation and Inverse Design of Metamaterials Based on a Deep Generative Model with Semi‐Supervised Learning Strategy (2019, background): “Unlike a very recently proposed generative model using generative adversarial network (GAN) [55] that needs a pre-trained simulator to guarantee the inverse design process, our model solves both the forward and inverse problem at the same time, and can be trained in an end-to-end manner.”
• Global optimization of dielectric metasurfaces using a physics-driven neural network (2019, background): “30,31 Another challenge is that many nanophotonic devices have complex curvilinear geometries and reside in a high dimensional design space, making it difficult for even the best networks to learn the nuanced relationship between device geometry and response.”
• Deep learning enabled inverse design in nanophotonics (2020, methodology): “[60], the authors also added a simulator network to approximate the optical properties from the generated design images (Figure 5A).”
• A Deep Learning Approach for Objective-Driven All-Dielectric Metasurface Design (2019, positive): “Compared to previously proposed DNN-based optical device design methods (17, 21-24), the PNN realized in this paper resolves difficulties associated with predicting resonant frequencies and achieves wideband phase prediction for the first time.”
• Physics-informed neural networks with hard constraints for inverse design (2021, methodology): “Al-ternatively, in unsupervised learning and reinforcement learning, neural networks are employed as generators to generate candidate designs [30, 18, 17, 52], which are then evaluated by numerical solvers.”