image: Inverse and forward design based on DL techniques. view more
Credit: OEA
A new publication from Opto-Electronic Advances; DOI 10.29026/oea.2022.210147 overviews the evaluation of metaphotonics induced by artificial intelligence and present a summary of the concepts of machine learning with some specific examples developed and demonstrated for metasystems and metasurfaces.
Advances in the field of artificial intelligence resulted in incorporation of these technologies into the process of scientific research and in the field of photonics. Such methods as machine learning and deep learning have become popular design tools for development of photonic devices. Design in this case implies prediction of a physical response of a given structure (forward design) as well as the reverse process of finding parameters of a structure required to provide a desired response (inverse design). While design procedures arguably remain the most widespread implementation of machine learning in photonics, novel applications begin to emerge leading to evolvement of a new research area of intelligent photonics.
The authors of this article overview recent advancements in the field of intelligent metaphotonics. This subfield of photonics is dedicated to intelligent systems and devices driven by optically induced electric and magnetic resonances achieved with the use of structured subwavelength nanoparticles (commonly called meta-atoms). The review covers machine learning for design of such devices as well as the other usages of machine learning, including classification tools, control systems and feedback mechanisms. Special attention is dedicated to potential practical applications, including solar cells, biosensors, or imagers. Avoiding diving into detailed description of methods, the authors tie together various aspects of machine learning and metaphotonics building a broad and coherent picture of their interplay, highlighting peculiarities of intelligent systems. Such novel concepts as self-adapting systems or intelligent biosensors are introduced and discussed in the context of future development of metaphotonics. Under self-adaptation the authors imply the ability of devices automatically tune their responses with the change of environmental conditions. One of the provided examples is a cloak adjusting itself to changes of frequency and angle of incidence of electromagnetic field, which allows preserving the functionality in a broad range of conditions. The concept of intelligent biosensors implies the usage of machine learning as a tool for classification of samples. The review also emphasises importance of metaphotonics for artificial intelligence demonstrating several examples of metasurfaces utilized as platforms for all-optical realization of machine learning algorithms.
For the broad audience the review may be interesting as an introduction to the field of intelligent metaphotonics as well as an overview of its state-of-art. No specific knowledge of computer science is required as the authors provide brief description of ideas behind ML and avoid technical details. The paper provides descriptive explanations of emerging concepts to make readers familiar with novel direction of development of photonics and provide references for consequent in-depth studies.
The authors of this article review the field of intelligent metaphotonics – area of science at the junction of artificial intelligence (AI) and metaphotonics.
AI rapidly becomes a part of work and daily life all over the world. Recent advances in this area have become possible largely due to the flourishing of such concept as machine learning (ML) representing a group of data-driven algorithms for learning ability of AI. It can be said that ML allows computer to learn from experience meaning that with the increased amount of input data a program may solve its pre-specified tasks better. In the same manner repetitions and trainings results in improved performance of humans with the exception that learning mechanisms and capabilities are quite different.
Science does not stay aside from modern trends and adopt advanced methods of computer science for problem solving and development of novel concepts. This article concentrates on application of ML in metaphotonics – a blossoming area of subwavelength photonics inspired by the physics of metamaterials. Initially introduced as a tool for forward and inverse design of photonic systems, ML has already evolved into something larger than just a method for design. The review provides examples of how ML was incorporated as an assisting tool for photonic sensors or feedback and control mechanism of self-adapting systems. Without any doubts ML rapidly becomes a powerful tool for research in the field of metaphotonics. Merging of these two areas led to development of a new area of research commonly named intelligent metaphotonics which implies metaphotonic systems designed or enhanced with ML or AI in general.
The article provides a brief introduction to AI concepts, throughout the paper ML is treated as a «black-box» providing a desired result without specific details on how exactly it was achieved. Such an approach allows the authors to focus on specific metaphotonic systems and their applications enabled by ML rather than on specific methods and algorithms. Also, in this way the paper may be interesting for readers not familiar with computer science. After the introduction, the paper covers ML-assisted design of nanoantennas which are building blocks of metaphotonic structures. Then, the focus shifts towards transformative metasurfaces and enhancement of their properties with ML. Special attention is dedicated to potential real-world applications, such as structural colours, LIDARs, or near-eye displays. Applications of metasurfaces as platforms for biosensing are presented in the separate section since in this field ML can be used not only as a design method but also as a tool for classification of the samples. This idea is demonstrated with examples of sensors for classification of SARS-CoV-2 and for monitoring biomolecule dynamics. Another class of applications involve self-adapting devices – metaphotonic systems which may automatically adapt to changes of the environment and adjust their responses. The provided examples include self-adapting microwave cloaks and imagers as well as metaphotonic systems used as computational platforms for ML. Finally, the authors provide an outlook covering emerging and potential trends in the field of intelligent metaphotonics assuming not only the influence of ML on photonics but also the vice versa importance of photonics for AI technologies.
Article reference: Krasikov S, Tranter A, Bogdanov A, Kivshar Y. Intelligent metaphotonics empowered by machine learning. Opto-Electron Adv 5, 210147 (2022). 10.29026/oea.2022.210147
Keywords: metaphotonics; machine learning; artificial intelligence
# # # # # #
The research was led by Professor Yuri Kivshar, who is a world leader in photonics and metamaterials, and one of the founders of the field of all-dielectric resonant metaphotonics governed by the physics of Mie resonances in dielectric nanoparticles with high refractive index.
# # # # # #
Opto-Electronic Advances (OEA) is a high-impact, open access, peer reviewed monthly SCI journal with an impact factor of 9.682 (Journals Citation Reports for IF 2020). Since its launch in March 2018, OEA has been indexed in SCI, EI, DOAJ, Scopus, CA and ICI databases over the time and expanded its Editorial Board to 36 members from 17 countries and regions (average h-index 49).
The journal is published by The Institute of Optics and Electronics, Chinese Academy of Sciences, aiming at providing a platform for researchers, academicians, professionals, practitioners, and students to impart and share knowledge in the form of high quality empirical and theoretical research papers covering the topics of optics, photonics and optoelectronics.
# # # # # #
More information: http://www.oejournal.org/oea
Editorial Board: http://www.oejournal.org/oea/editorialboard/list
All issues available in the online archive (http://www.oejournal.org/oea/archive).
Submissions to OEA may be made using ScholarOne (https://mc03.manuscriptcentral.com/oea).
ISSN: 2096-4579
CN: 51-1781/TN
Contact Us: oea@ioe.ac.cn
Twitter: @OptoElectronAdv (https://twitter.com/OptoElectronAdv?lang=en)
WeChat: OE_Journal
# # # # # #
Journal
Opto-Electronic Advances