CDS&E: Elucidating and Controlling the Spectral, Spatial and Temporal Responses of Plasmonic Nanostructures based on a Data-Driven Approach

非技术合成等离子体纳米结构是由贵金属(如金和银)制成的微小颗粒。它们可以将光线集中在极小的维度上，并显著增强光线强度。等离子体纳米结构的独特性质已经产生了许多应用，包括太阳能收集、生物医学传感、诊断和治疗。该项目的目标是建立一个基于深度学习的框架，通过融合理论、计算、深度学习算法和实验来阐明和控制等离子体纳米结构的物理性质。最近的工作表明，深度学习可以通过大量的数据发现光子结构与其性质之间高度复杂和非直观的关系，克服了传统解析和数值方法的局限性。为了进一步推进这一新兴领域，PI将通过考虑更多自由度来增强深度学习模型的能力，如电场和磁场分布、随时间变化的响应以及等离子体纳米结构中的多物理过程。深度学习模型的预测将直接通过实验来验证，实验将提供重要的反馈和额外的数据来提高模型的能力。该奖项还支持一项全面的教育计划，其中将包括7-12年级、本科生和研究生水平的创新活动。将特别努力吸引和教育未被充分代表的民族/种族和性别群体的学生，并拓宽他们在光子学、材料科学、应用物理和人工智能方面的知识。部分研究成果将用于开发新的课程材料，向东北大学的学生介绍该领域的最新发展。PI还将创建一个新的推广活动，在该活动中，当地高中生将学习等离子体纳米结构的光学性质。技术总结该奖项支持旨在通过利用先进的深度学习技术并以等离子体纳米结构为平台，有效利用光谱、空间和时间域中的光-物质相互作用的计算和实验研究。该项目包括三个研究方向：(1)开发一种通用的方法来模拟复杂等离子体纳米结构的电场和磁场的空间分布，这是在用户定义的波长上设计非线性光学效应的关键；(2)通过集成深度学习、傅里叶变换和控制方程来研究热电子诱导的时间响应和由此产生的等离子体纳米结构的瞬时多物理过程；以及(3)制备具有规范和自由形状的等离子体纳米结构，并利用先进的显微镜和光谱技术对其进行实验表征。由深度学习实现的以数据为中心的框架将有助于揭示等离子体纳米结构和其他光子设计背后的物理原理。它将提供对一系列基本问题的深刻见解，如拓扑、对称性和非平衡动力学，具有良好的空间和时间分辨率。该奖项还支持一项全面的教育计划，其中将包括7-12年级、本科生和研究生水平的创新活动。将特别努力吸引和教育未被充分代表的民族/种族和性别群体的学生，并拓宽他们在光子学、材料科学、应用物理和人工智能方面的知识。部分研究成果将用于开发新的课程材料，向东北大学的学生介绍该领域的最新发展。PI还将创建一项新的推广活动，让当地高中生了解等离子体纳米结构的光学性质。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

NONTECHNICAL SUMMARYPlasmonic nanostructures are tiny particles made of noble metals, such as gold and silver. They can concentrate light into extremely small dimensions and significantly enhance the light intensity. The unique properties of plasmonic nanostructures have produced many applications, including solar energy harvesting, biomedical sensing, diagnostics and therapy. The objective of this project is to establish a framework based on deep learning, a subset of artificial intelligence, to elucidate and control the physical properties of plasmonic nanostructures by fusing theory, computation, deep learning algorithms, and experiments. Recent work has demonstrated that deep learning can discover the highly complicated and non-intuitive relationships between photonic structures and their properties through extensive data, overcoming the limitations of conventional analytical and numerical methods. To further advance this emergent field, the PI will enhance the capability of deep learning models by considering more degrees of freedom, such as electric and magnetic field distributions, time-dependent responses, and multi-physics processes in plasmonic nanostructures. The predictions of the deep learning models will be directly validated by experiments, which will provide important feedback and additional data to improve the model capability.This award also supports a comprehensive education plan that will include innovative activities at the Grade 7-12, undergraduate, and graduate levels. Special efforts will be made to attract and educate students from underrepresented ethnic/racial and gender groups, and broaden their knowledge in photonics, materials science, applied physics, and artificial intelligence. Part of the research findings will be used to develop new course materials to introduce the latest development in the fields to the students at Northeastern University. The PI will also create a new outreach activity, in which local high-school students will learn about optical properties of plasmonic nanostructures.TECHNICAL SUMMARYThis award supports computational and experimental research aimed at effectively harnessing light-matter interactions in the spectral, spatial, and temporal domains by leveraging advanced deep learning techniques and using plasmonic nanostructures as the platform. The project consists of three research thrusts: (1) developing a general approach to model the spatial distribution of the electric and magnetic fields of complex plasmonic nanostructures with high efficiency, accuracy and fidelity, which is critical to engineer nonlinear optical effects at user-defined wavelengths; (2) investigating the hot-electron-induced temporal responses and the resulting transient multi-physics processes of plasmonic nanostructures by integrating deep learning, Fourier transform and governing equations; and (3) fabricating plasmonic nanostructures with canonical and freeform shapes, and experimentally characterizing them using advanced microscopy and spectroscopy techniques. The data-centric framework enabled by deep learning will help to uncover the physics behind plasmonic nanostructures and other photonic designs. It will provide deep insights into a series of fundamental problems, such as topology, symmetry, and non-equilibrium dynamics, with fine spatial and temporal resolution.This award also supports a comprehensive education plan that will include innovative activities at the Grade 7-12, undergraduate, and graduate levels. Special efforts will be made to attract and educate students from underrepresented ethnic/racial and gender groups, and broaden their knowledge in photonics, materials science, applied physics, and artificial intelligence. Part of the research findings will be used to develop new course materials to introduce the latest development in the fields to the students at Northeastern University. The PI will also create a new outreach activity, in which local high-school students will learn about optical properties of plasmonic nanostructures.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.