Adjoint-Assisted Shape Optimization of Microlenses for CMOS Image Sensors DOI Creative Commons

Rishad Arfin,

Jens Niegemann,

Dylan McGuire

и другие.

Sensors, Год журнала: 2024, Номер 24(23), С. 7693 - 7693

Опубликована: Ноя. 30, 2024

Recently, there have been significant developments in the designs of CMOS image sensors to achieve high-resolution sensing capabilities. One fundamental factors determining sensor's ability capture images is its efficiency focusing visible light onto photosensitive region submicron scale. In most imaging technologies, this typically achieved through microlenses. Light collection under diverse conditions can be significantly improved efficient design While optimization microlenses appears imperative, achieving for high-density pixels various remains a challenge. Therefore, systematic approach required accelerate development with enhanced optical performance. paper, we present an optimize shape adjoint sensitivity analysis (ASA). A novel figure merit (FOM) developed and incorporated into process enhance collection. The gradient FOM computed iteratively using two field simulations only. functionality robustness framework are thoroughly evaluated. Furthermore, performance optimized compared that conventional adjoint-assisted presented here further used develop devices perform manipulation such as concentrating, bending, or dispersing compact systems.

Язык: Английский

Data-efficient prediction of OLED optical properties enabled by transfer learning DOI Creative Commons

Jeong Min Shin,

Sanmun Kim,

Sergey G. Menabde

и другие.

Nanophotonics, Год журнала: 2025, Номер 14(8), С. 1091 - 1099

Опубликована: Фев. 8, 2025

Abstract It has long been desired to enable global structural optimization of organic light-emitting diodes (OLEDs) for maximal light extraction. The most critical obstacles achieving this goal are time-consuming optical simulations and discrepancies between simulation experiment. In work, by leveraging transfer learning, we demonstrate that fast reliable prediction OLED properties is possible with several times higher data efficiency compared previously demonstrated surrogate solvers based on artificial neural networks. Once a network trained base structure, it can be transferred predict the modified structures additional layers relatively small number training samples. Moreover, that, only few tenths experimental sets, accurately measurements OLEDs, which often differ from results due fabrication measurement errors. This enabled transferring pre-trained network, built large amount simulated data, new capable correcting systematic errors in Our work proposes practical approach designing optimizing design parameters achieve high efficiency.

Язык: Английский

Процитировано

1
Revolutionary Integration of Artificial Intelligence with Meta-Optics-Focus on Metalenses for Imaging DOI Creative Commons
Nikolay L. Kazanskiy, Svetlana N. Khonina, Ivan Oseledets

и другие.

Technologies, Год журнала: 2024, Номер 12(9), С. 143 - 143

Опубликована: Авг. 28, 2024

Artificial intelligence (AI) significantly enhances the development of Meta-Optics (MOs), which encompasses advanced optical components like metalenses and metasurfaces designed to manipulate light at nanoscale. The intricate design these requires sophisticated modeling optimization achieve precise control over behavior, tasks for AI is exceptionally well-suited. Machine learning (ML) algorithms can analyze extensive datasets simulate numerous variations identify most effective configurations, drastically speeding up process. also enables adaptive MOs that dynamically adjust changing imaging conditions, improving performance in real-time. This results superior image quality, higher resolution, new functionalities across various applications, including microscopy, medical diagnostics, consumer electronics. combination with thus epitomizes a transformative advancement, pushing boundaries what possible technology. In this review, we explored latest advancements AI-powered applications.

Язык: Английский

Процитировано

5
Advances in machine learning optimization for classical and quantum photonics DOI
María José Sánchez, Christer Everly, P. A. Postigo

и другие.

Journal of the Optical Society of America B, Год журнала: 2024, Номер 41(2), С. A177 - A177

Опубликована: Янв. 8, 2024

The development and optimization of photonic devices various other nanostructure electromagnetic present a computationally intensive task. Much relies on finite-difference time-domain or finite element analysis simulations, which can become very demanding for finely detailed structures dramatically reduce the available space. In recent years, inverse design machine learning (ML) techniques have been successfully applied to realize previously unexplored spaces quantum devices. this review, results using conventional methods, such as adjoint method particle swarm, are examined along with ML convolutional neural networks, Bayesian optimizations deep learning, reinforcement in context new applications photonics photonics.

Язык: Английский

Процитировано

4
Highly Efficient and achromatic mid-infrared silicon nitride meta-lenses DOI Creative Commons

Abdullah Maher,

Mohamed A. Swillam

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 15, 2025

Abstract Inverse design with topology optimization considers a promising methodology for discovering new optimized photonic structure that enables to break the limitations of forward or traditional especially meta-structure. This work presents high efficiency mid infra-red imaging photonics element along wavelengths band starts from 2 5 µm based on silicon nitride material structures. The first two designs are broadband focusing and reflective meta-lens under very numerical aperture condition (NA = 0.9). modeled by inverse problem Kreisselmeier–Steinhauser (k–s) aggregation objective function, while final is depended novel double function can target bifocal points wavelength producing achromatic multi-focal apertures ( $$N{A}_{1} 0.9, \, N{A}_{2}=0.88$$ ).

Язык: Английский

Процитировано

0
Exploring AI in Metasurface Structures with Forward and Inverse Design DOI Creative Commons

Guantai Yang,

Qingxiong Xiao,

Zhilin Zhang

и другие.

iScience, Год журнала: 2025, Номер 28(3), С. 111995 - 111995

Опубликована: Фев. 15, 2025

As an artificially manufactured planar device, a metasurface structure can produce unusual electromagnetic responses by harnessing four basic characteristics of the light wave. Traditional design processes rely on numerical algorithms combined with parameter optimization. However, such methods are often time-consuming and struggle to match actual responses. This paper aims give unique perspective classify artificial intelligence(AI)-enabled design, dividing it into forward inverse designs according mapping relationship between variables performance. Forward driven intelligent algorithms; neural networks one principal ways realize reverse design. reviews recent progress in AI-enabled examining principles, advantages, potential applications. A rich content detailed comparison help build holistic understanding Moreover, authors believe that this systematic review will pave way for future research selection practical

Язык: Английский

Процитировано

0
Technological Advancements in Photonics and Electronics: The Machine Learning Perspective DOI

Shital Tank,

Priyanka Mishra,

Mahuya Bandyopadhyay

и другие.

Studies in Infrastructure and Control, Год журнала: 2025, Номер unknown, С. 89 - 101

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

0
Synergy between AI and Optical Metasurfaces: A Critical Overview of Recent Advances DOI Creative Commons
Zoran Jakšić

Photonics, Год журнала: 2024, Номер 11(5), С. 442 - 442

Опубликована: Май 9, 2024

The interplay between two paradigms, artificial intelligence (AI) and optical metasurfaces, nowadays appears obvious unavoidable. AI is permeating literally all facets of human activity, from science arts to everyday life. On the other hand, metasurfaces offer diverse sophisticated multifunctionalities, many which appeared impossible only a short time ago. use for optimization general approach that has become ubiquitous. However, here we are witnessing two-way process—AI improving but some also AI. helps design, analyze utilize while ensure creation all-optical chips. This ensures positive feedback where each enhances one: this may well be revolution in making. A vast number publications already cover either first or second direction; modest includes both. an attempt make reader-friendly critical overview emerging synergy. It succinctly reviews research trends, stressing most recent findings. Then, it considers possible future developments challenges. author hopes broad interdisciplinary will useful both dedicated experts scholarly audience.

Язык: Английский

Процитировано

3
The dynamic integration of computational approaches and machine learning for cutting-edge solutions in photonics DOI

Sakshi Gulia,

Mirza Tanweer Ahmad Beig,

Rajiv Vatsa

и другие.

Deleted Journal, Год журнала: 2024, Номер 245(1)

Опубликована: Май 9, 2024

This comprehensive study delves into the transformative evolution of photonic feature prediction and design, where traditional methods, deeply rooted in theory-driven computational approaches, have shaped our understanding optical phenomena advanced structures. Integrating machine learning (ML) photonics marks a fundamental departure from conventional predictive modeling, driven by acknowledgment its vast potential to deliver ingenious solutions, optimize designs, accelerate advancement cutting-edge technologies. The article introduces practical application learning, specifically regression, address engineering problems. focal point is hexagonal crystal fiber (PCF), an important device with crucial input parameters such as wavelength, diameter, pitch guiding analysis. hands-on ML showcases adaptability techniques. It underscores pivotal role creating robust dataset foundational step for effective model training problem-solving systems. synergy between models data-driven approaches explored, revealing promising era unlocking novel insights driving innovation photonics, features devices. shift towards methodologies addresses prevailing limitations methods when navigating intricate complexities inherent Research dynamic interplay established theories emerging poised uncover insights, ultimately field efficiently solving complex systems deploying effectively optimized neural networks predict specific outputs given inputs.

Язык: Английский

Процитировано

2
Beyond dichotomy: unveiling mode behavior in plasmonic nanodisks DOI
Ayda Aray, Saeed Ghavami Sabouri,

Sara Sadat Ghaffari-Oskooei

и другие.

Applied Optics, Год журнала: 2024, Номер 63(21), С. 5738 - 5738

Опубликована: Июль 1, 2024

In our study, we investigate the resonance modes of plasmonic nanodisks through numerical simulations and theoretical analysis. These tiny structures exhibit fascinating behavior, but relying solely on mode localization is not sufficient to classify their supported as or dielectric. Our goal address this challenge by introducing a robust method for identifying each mode’s true nature. Moreover, analysis field distribution, introduce, knowledge, novel metric designed application in inverse problems within realm machine learning. This serves tool optimizing performance photonic devices.

Язык: Английский

Процитировано

1
Deep-learning empowered unique and rapid optimization of meta-absorbers for solar thermophotovoltaics DOI Creative Commons

Sadia Noureen,

Sumbel Ijaz, Isma Javed

и другие.

Optical Materials Express, Год журнала: 2024, Номер 14(4), С. 1025 - 1025

Опубликована: Март 14, 2024

Optical nano-structure designs usually employ computationally expensive and time-intensive electromagnetic (EM) simulations that call for resorting to modern-day data-oriented methods, making design robust quicker. A unique dataset hybrid image processing model combining a CNN with gated recurrent units is presented foresee the EM absorption response of photonic nano-structures. An inverse also discussed predict optimum geometry dimensions meta-absorbers. Mean-squared error order 10 −3 an accuracy 99% achieved trained models, average prediction time DL models around 98% faster than simulations. This idea strengthens proposition efficient DL-based solutions can substitute traditional methods designing nano-optical structures.

Язык: Английский

Процитировано

0