Review
Chemistry, Multidisciplinary
Zhaocheng Liu, Dayu Zhu, Lakshmi Raju, Wenshan Cai
Summary: Machine learning, as an algorithm study for automated prediction and decision-making based on complex data, has become an indispensable tool in artificial intelligence research, leading to significant progress in scientific research fields like quantum physics, organic chemistry, and medical imaging. Recently adopted in photonics and optics research, machine learning shows potential for addressing the inverse design problem and has been instrumental in advancing photonic design strategies in recent years, particularly focusing on deep learning methods for high degrees-of-freedom structure design.
Article
Materials Science, Multidisciplinary
Christopher Yeung, Ryan Tsai, Benjamin Pham, Brian King, Yusaku Kawagoe, David Ho, Julia Liang, Mark W. Knight, Aaswath P. Raman
Summary: This study presents a global deep learning-based inverse design framework for optimizing photonic metasurfaces by training a neural network on images encoded with material and structural parameters. The network can identify effective metasurface designs based on target absorption spectra and generate multiple design variants with nearly identical absorption spectra.
ADVANCED OPTICAL MATERIALS
(2021)
Article
Materials Science, Ceramics
Xiang Xu, Jingyi Hu
Summary: Metallic glass has attracted attention due to its unique properties, but the complex composition design space poses challenges for traditional experimental methods. This paper proposes a novel approach using a generative adversarial network (GAN) to generate hypothetical metallic glass compositions quickly. The GAN-generated samples were evaluated for validity, novelty, and uniqueness. Results show high validity rates and demonstrate the novelty and uniqueness of the generated samples through distribution comparison. The GAN model is expected to improve sampling efficiency and shorten the development cycle of metallic glass.
JOURNAL OF NON-CRYSTALLINE SOLIDS
(2023)
Article
Engineering, Electrical & Electronic
Zheming Gu, Da Li, Yunlong Wu, Yudi Fan, Chengting Yu, Hongsheng Chen, Er-Ping Li
Summary: Recently, artificial neural networks (ANNs) have shown great potential in frequency-selective surface (FSS) inverse design. However, the problem of nonunique mapping between inputs and outputs cannot be easily solved by traditional ANNs framework. In this study, we propose deploying generative models as a solution for the first time, and present two approaches with a novel model based on conditional generative adversarial network (cGAN) to achieve inverse design from given indexes to FSS physical dimensions. The proposed method allows immediate obtaining of FSS designs that meet industrial demands without complex neural network processing or repeated iterations, and has been validated in closed-loop simulations and experiments for designing complex FSS structures with desired electromagnetic responses using deep neural networks.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION
(2023)
Article
Nanoscience & Nanotechnology
A. Nikulin, I. Zisman, M. Eich, A. Yu. Petrov, A. Itin
Summary: The article introduces several approaches to predict and optimize the properties of photonic crystals using ML methods, including the application of symmetry-aware augmentations and hybrid ML-solver approaches to improve the performance of predictive models, as well as the use of VAEs combined with predictor models to generate photonic structures.
PHOTONICS AND NANOSTRUCTURES-FUNDAMENTALS AND APPLICATIONS
(2022)
Article
Nanoscience & Nanotechnology
Christopher Yeung, Benjamin Pham, Ryan Tsai, Katherine T. Fountaine, Aaswath P. Raman
Summary: In recent years, hybrid design strategies combining machine learning with electromagnetic optimization algorithms have been used for the inverse design of photonic structures and devices. These methods can rapidly identify optimal solutions and reduce computational costs. However, there is a need to optimize across both materials and geometries in a single integrated environment.
Article
Multidisciplinary Sciences
Niklas W. A. Gebauer, Michael Gastegger, Stefaan S. P. Hessmann, Klaus-Robert Mueller, Kristof T. Schuett
Summary: The authors propose a conditional generative neural network for the inverse design of 3d molecular structures. This approach allows targeted sampling of novel molecules with specified chemical and structural properties, even in domains with sparse reference calculations.
NATURE COMMUNICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Jiyoung Jung, Kundo Park, Byungjin Cho, Jinkyoo Park, Seunghwa Ryu
Summary: In this paper, two systematic data-driven optimization frameworks for the injection molding process are proposed, using multi-objective Bayesian optimization and constrained generative inverse design network frameworks. These methods can efficiently obtain optimal process parameters and are demonstrated to be applicable in the manufacturing process of a door trim part.
JOURNAL OF INTELLIGENT MANUFACTURING
(2023)
Article
Engineering, Electrical & Electronic
Pakshal Bohra, Thanh-an Pham, Jonathan Dong, Michael Unser
Summary: This study presents a Bayesian reconstruction framework for nonlinear imaging models, where the prior knowledge on the image is specified through a deep generative model. The authors develop a tractable posterior-sampling scheme based on the Metropolis-adjusted Langevin algorithm for the class of nonlinear inverse problems. The advantages of this framework are illustrated through its application to various nonlinear imaging modalities, such as phase retrieval and optical diffraction tomography.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2022)
Article
Materials Science, Multidisciplinary
Teng Long, Yixuan Zhang, Nuno M. Fortunato, Chen Shen, Mian Dai, Hongbin Zhang
Summary: We developed an inverse design framework that enables automated generation of stable multicomponent crystal structures and discovered unreported crystal structures through analysis. This method provides convenience for inverse design of multicomponent materials with possible multi-objective optimization.
Article
Chemistry, Physical
Jorge-Alberto Peralta-Angeles, Jorge-Alejandro Reyes-Esqueda
Summary: This study presents an analytical and numerical study of hybrid photonic-plasmonic crystals. The proposed theoretical model describes a system consisting of a dielectric photonic crystal on a metallic thin film. By analyzing different crystal structures, the characteristics of photonic bandgaps are determined. Artificial neural networks are trained using the analytical model to predict the center and width of bandgaps, achieving a prediction accuracy of over 95%. This research provides a useful tool for tuning the optical properties of hybrid photonic-plasmonic crystals, with potential applications in various areas.
Article
Engineering, Electrical & Electronic
Xin Tu, Wansheng Xie, Zhenmin Chen, Ming-Feng Ge, Tianye Huang, Chaolong Song, H. Y. Fu
Summary: By utilizing deep neural networks and finite-difference time-domain solver, efficient design of silicon photonic devices can be achieved; the study shows that both forward and inverse design methods can result in high prediction accuracy for the coupler, reaching up to 91.7%.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
M. A. G. Duff, N. D. F. Campbell, M. J. Ehrhardt
Summary: Deep neural network approaches have achieved impressive results in inverse imaging problems. This survey paper explores the usage of generative models in a variational regularization approach for solving inverse problems. The authors evaluate the quality of generative models and propose a set of criteria for assessing them, as well as testing different generative models and regularizers in numerical experiments. The findings suggest that allowing small deviations from the range of the generator produces more consistent results in solving inverse problems.
JOURNAL OF MATHEMATICAL IMAGING AND VISION
(2023)
Article
Nanoscience & Nanotechnology
Renjie Li, Ceyao Zhang, Wentao Xie, Yuanhao Gong, Feilong Ding, Hui Dai, Zihan Chen, Feng Yin, Zhaoyu Zhang
Summary: Photonics inverse design is traditionally done by human experts, which is labor-intensive, slow, and sub-optimal. This study proposes L2DO, a reinforcement learning-based method that autonomously designs nanophotonic laser cavities and outperforms human experts by over 2 orders of magnitude. L2DO can bypass the shortage of training data, non-unique solutions, and optimization limitations of traditional supervised or semi-supervised learning approaches.
Article
Materials Science, Multidisciplinary
Chao Ma, Gang Li, Longhui Qin, Weicheng Huang, Hongrui Zhang, Wenfeng Liu, Tianyu Dong, Sheng-Tao Li
Summary: Flexible micro-pyramidal capacitive pressure sensors offer high tunability and have diverse applications in advanced healthcare, prosthetics, and smart robots. Analytical models based on micro-pyramidal electrodes and dielectrics have been proposed and verified through finite element simulations and experimental results to predict pressure responses and analyze device design parameters. Machine learning techniques, such as neural networks, are utilized for approximating pressure responses and optimizing sensor parameters for customized performance.
ADVANCED MATERIALS TECHNOLOGIES
(2021)
Article
Multidisciplinary Sciences
P. A. D. Goncalves, Thomas Christensen, Nicholas Rivera, Antti-Pekka Jauho, N. Asger Mortensen, Marin Soljacic
NATURE COMMUNICATIONS
(2020)
Article
Nanoscience & Nanotechnology
Sophie Fisher, Charles Roques-Carmes, Nicholas Rivera, Liang Jie Wong, Ido Kaminer, Marin Soljacic
Article
Materials Science, Multidisciplinary
Ravishankar Sundararaman, Thomas Christensen, Yuan Ping, Nicholas Rivera, John D. Joannopoulos, Marin Soljacic, Prineha Narang
PHYSICAL REVIEW MATERIALS
(2020)
Article
Multidisciplinary Sciences
Liang Jie Wong, Nicholas Rivera, Chitraang Murdia, Thomas Christensen, John D. Joannopoulos, Marin Soljacic, Ido Kaminer
Summary: The researchers demonstrate that controlling radiation emitted by individual electrons is achievable by shaping the electron wavepacket, leading to examples of collimated and monochromatic X-ray emission from specially shaped electrons. This ability opens up additional avenues of control in phenomena ranging from optical excitation in electron microscopy to free electron lasing, highlighting the potential for tailored general QED processes in the quantum regime.
NATURE COMMUNICATIONS
(2021)
Article
Multidisciplinary Sciences
P. A. D. Goncalves, Thomas Christensen, Nuno M. R. Peres, Antti-Pekka Jauho, Itai Epstein, Frank H. L. Koppens, Marin Soljacic, N. Asger Mortensen
Summary: Understanding the quantum response of materials is crucial for designing light-matter interactions at the nanoscale. Graphene plasmons can be utilized to probe the quantum surface-response of metals with subnanometer resolution. This study demonstrates a promising approach for inferring metallic quantum response from measurements by using acoustic graphene plasmons.
NATURE COMMUNICATIONS
(2021)
Article
Multidisciplinary Sciences
Charlotte Loh, Thomas Christensen, Rumen Dangovski, Samuel Kim, Marin Soljacic
Summary: Deep learning techniques often require a large amount of training data, which can be challenging in the case of scarce datasets. This study proposes a framework that combines contrastive and transfer learning to reduce the data requirements for training while maintaining prediction accuracy. By utilizing auxiliary information sources, such as unlabeled data, prior knowledge, and surrogate data, the proposed framework consistently achieves significant reductions in the number of labels needed for accurate predictions.
NATURE COMMUNICATIONS
(2022)
Article
Physics, Multidisciplinary
Thomas Christensen, Hoi Chun Po, John D. Joannopoulos, Marin Soljacic
Summary: This paper addresses the opening of the fundamental band gap in three-dimensional photonic crystals and reveals the influence of topology on this gap. By introducing auxiliary longitudinal modes, the authors overcome the singularity caused by symmetry and systematically study the topology of the minimal fundamental gaps.
Article
Chemistry, Multidisciplinary
Andrew Ma, Yang Zhang, Thomas Christensen, Hoi Chu Po, Li Jing, Liang Fu, Marin Soljacic
Summary: By utilizing machine learning, we have developed a simple chemical rule that accurately diagnoses whether a material is topological or not, based solely on its chemical formula. We have also established a high-throughput procedure for discovering topological materials using this heuristic rule, followed by ab initio validation. This approach has led to the discovery of new topological materials that cannot be identified using symmetry indicators, some of which show promise for experimental observation.
Article
Nanoscience & Nanotechnology
Samuel Kim, Thomas Christensen, Steven G. Johnson, Marin Soljacic
Summary: Topological photonic crystals have attracted attention for their unique ability to manipulate and guide light. A combined global and local optimization framework has been proposed to optimize these crystals, using a flexible symmetry-constrained level-set parametrization and standard gradient-free optimization algorithms. This framework can be applied to any symmetry-identifiable band topology, enabling the automated discovery of novel designs without prior examples or knowledge.
Article
Materials Science, Multidisciplinary
Andre Grossi e Fonseca, Thomas Christensen, John D. Joannopoulos, Marin Soljacic
Summary: We introduce a general mechanism for obtaining Weyl points in a stack of two-dimensional quasicrystals, and provide an example using a Penrose quasicrystal. This mechanism can be extended to any stack of aperiodic layers, and we also uncover an analog of Fermi-Bragg arcs.
Proceedings Paper
Engineering, Electrical & Electronic
P. A. D. Goncalves, T. Christensen, N. M. R. Peres, P. A. Jauho, I Epstein, F. H. L. Koppens, M. Soljacic, N. A. Mortensen
2021 CONFERENCE ON LASERS AND ELECTRO-OPTICS EUROPE & EUROPEAN QUANTUM ELECTRONICS CONFERENCE (CLEO/EUROPE-EQEC)
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Yi Yang, Di Zhu, Wei Ya, Akshay Agrawal, Mengjie Zheng, John D. Joannopoulos, Philippe Lalanne, Thomas Christensen, Karl Berggren, Marin Soljacic
2020 CONFERENCE ON LASERS AND ELECTRO-OPTICS (CLEO)
(2020)
Proceedings Paper
Engineering, Electrical & Electronic
Liang Jie Wong, Nicholas Rivera, Chitraang Murdia, Thomas Christensen, John D. Joannopoulos, Marin Soljacic, Ido Kaminer
2020 CONFERENCE ON LASERS AND ELECTRO-OPTICS (CLEO)
(2020)