ERI: Sub-diffractive Optical Trapping Enabled by Deep-Learning-Assisted Metasurface Design

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。光与不同材料和结构之间复杂的相互作用使得许多日常应用成为可能：用太阳能电池发电，用光学显微镜放大非常小的物体，或者通过光纤传输大量数据。除了这些众所周知的电磁（EM）相互作用之外，光还可以对物体施加压力（即力），如果控制得当，可以允许非常小的粒子移动和定位。例如，纳米粒子、细胞和分子比人类头发的大小小10到1000倍，因此需要按比例精确控制施加在它们身上的力，以实现所需的运动或定位。这直接取决于光能在多大程度上被限制在小范围内，以及塑造粒子周围能量局部分布的能力。传统的光捕获装置（称为光镊子）是操纵微米级物体（如细胞等百万分之一米大小）的好选择，但不能产生纳米级限制（如蛋白质等十亿分之一米大小），因此不适合操纵单个纳米级物体。为了规避这些限制并实现光学纳米捕获的新自由度，该项目旨在利用深度学习算法在精细图案表面（称为超表面）附近创建最佳能量分布以捕获纳米颗粒。这些表面由低损耗的介电材料制成，其外形设计用于产生纳米级的光约束，并有效地捕获不同大小和形状的粒子。此外，PI还参与教育和推广活动，包括指导本科生研究和为K-12和大学生组织研讨会，重点是促进科学、技术、工程和数学（STEM），并向学生介绍计算机科学在其他科学领域的应用。技术描述：该项目旨在利用超表面的波形特性来设计一种新型的紧凑纳米结构，用于亚衍射光学捕获，能够产生所需的捕获势，并为选择的粒子量身定制巨大的光学力。传统的光学捕获方法不容易实现纳米尺度的光聚焦，并且在实现小于100纳米的粒子的有效光学纳米捕获方面面临主要障碍。这些障碍包括对高强度激光产生局部力的要求，这反过来可能导致样品的光热损伤，以及由于大聚焦区域而无法区分和捕获单个纳米颗粒。通过利用元表面提供的大设计空间，利用人工神经网络捕获高度非线性和复杂的粒子场相互作用，可以解决这些限制。该项目的具体目标包括：(i)在超表面附近产生高度受限的捕获势，（ii）根据不同颗粒形状/大小定制捕获势的能力，（iii）在表面的不同方向和位置产生局部排斥和吸引力，以及（iv）在防止高激光功率造成永久性光热损伤的同时，创造适合亚波长纳米颗粒的增强光力。此外，PI从事研究指导和专题讨论会的组织，重点是促进STEM和向学生介绍深度学习在光子结构设计中的意义。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).The complex interactions of light with different materials and structures enable several everyday applications: generating electricity with solar cells, magnifying very small objects with light microscopes, or transmitting large amounts of data through optical fibers. Beyond these well-known electromagnetic (EM) interactions, light can also exert pressure (i.e., force) on objects which, if properly controlled, can allow for movement and positioning of very small particles. For instance, nanoparticles, cells, and molecules are 10s to 1000s of times smaller than the size of the human hair and thus require a proportionally accurate control of the force exerted on them to enable the desired movement or positioning. This directly depends on how well light can be confined into small areas as well as the ability to shape the local distribution of the energy around the particles. Conventional optical trapping devices (called optical tweezers) are a good candidate for manipulating micron-sized objects (sizes in the order of one millionth of a meter such as cells) but cannot create nanoscale confinement (sizes in the order of one billionth of a meter such as proteins), thus are not suitable for manipulating individual nanoscale objects. To circumvent these limitations and enable new degrees of freedom in optical nano-trapping, this project is designed to utilize deep learning algorithms for creating optimal energy distributions near a finely patterned surface (called metasurfaces) to trap nanoparticles. These surfaces are made of low-loss dielectric materials whose profile is designed to create nanoscale light confinement and to efficiently trap particles with different sizes and shapes. In addition, the PI engages in educational and outreach activities including undergraduate research mentoring and workshop organization for K-12 and college students, focusing on promoting science, technology, engineering, and mathematics (STEM), and introducing students to applications of computer science in other scientific fields.Technical description: This project aims to leverage the wave-shaping properties of metasurfaces to design a novel class of compact nanostructures for sub-diffractive optical trapping, able to generate desired trapping potentials with giant optical forces that are tailored for the particles of choice. Conventional approaches for optical trapping do not readily allow nanoscale focusing of light and face major barriers in achieving effective optical nano-trapping for particles smaller than 100 nm. These barriers include requirement for high intensity lasers to generate local force which in return may cause photothermal damage to the sample as well as failure to distinguish and trap single nanoparticles due to large focal area. These limitations are addressed by using artificial neural networks to capture the highly nonlinear and complex particle-field interactions while leveraging the large design space offered by metasurfaces. The specific objectives of the project include: (i) generating highly confined trapping potentials near metasurfaces, (ii) ability to tailor the trapping potential for various particle shapes/sizes, (iii) creating local repulsive and attractive forces in different directions and locations on the surface, and (iv) creating enhanced optical forces suitable for subwavelength nanoparticles while preventing permanent photothermal damage resulted from high laser power. In addition, PI engages in research mentoring and organization of symposia focused on promoting STEM and introducing students to implications of deep learning in the design of photonic structures.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.