Advances in the field of plasmonics, that is, nanophotonics based on optical properties of metal nanostructures, paved the way for the development of ultrasensitive biological sensors and other devices whose operating principles are based on localization of an electromagnetic field at the nanometer scale. However, high dissipative losses of metal nanostructures limit their performance in many modern areas, including metasurfaces, metamaterials, and optical interconnections, which required the development of new devices that combine them with high refractive index dielectric nanoparticles. Resulting metal-dielectric (hybrid) nanostructures demonstrated many superior properties from the point of view of practical application, including moderate dissipative losses, resonant optical magnetic response, strong nonlinear optical properties, which made the development in this field the vanguard of the modern light science. This review is devoted to the current state of theoretical and experimental studies of hybrid metal-dielectric nanoantennas and nanostructures based on them, capable of selective scattering light waves, amplifying and transmitting optical signals in the desired direction, controlling the propagation of such signals, and generating optical harmonics.
Deep Neural Network Inverse Design of Integrated Nanophotonic Devices
预测人工结构材料的物理响应对于科学和工程应用特别感兴趣。在这里，我们使用深度学习来预测人工设计的纳米光子器件的光学响应。除了预测任何给定拓扑结构的传输响应的正向近似外，这种方法还允许我们逆近似设计目标光学响应。我们的深度神经网络（DNN）可以在几分之一秒内设计出紧凑的（2.6x2.6 μm2）基于绝缘体上硅（SOI）的1 X 2功率分配器，具有各种目标分束比。该模型经过训练，以最大限度地减少反射（小于20 dB），同时实现最大传输效率（高于90%）和目标分裂规格。这种方法为快速设计依赖于复杂纳米结构的集成光子器件铺平了道路。
