| Optical Metasurfaces are artificial micro and nano structures with hundreds of nanometers thickness,the sub-wavelength scale"superatomic"unit structures can interact uniquely with electromagnetic fields to modulate the properties of light,such as controlling the propagation direction of light beams,achieving wavefront shaping（focusing）and enhancing the near field,and therefore have important applications in sensing,imaging and communication.How to design customized supersurface structures quickly and efficiently according to the requirements has been a focus of research in this field.On the other hand,the fast and accurate identification of supersurface structure parameters as well as the unification of theoretical design and experimental results is also an urgent challenge.In this paper,we will focus on two directions:recognition of plasmonic nanostructure by neural-networks-assisted spectroscopy and the intelligent design of dielectric metasurface supporting continuum-bound states.The lateral geometry and material property of plasmonic nanostructures are critical parameters for tailoring their optical resonance for sensing applications.While lateral geometry can be easily observed by a scanning electron microscope or an atomic force microscope,characterizing materials properties of plasmonic devices is not straightforward and requires delicate examination of material composition,cross-sectional thickness,and refractive index.In this study,a deep neural network is adopted to characterize these parameters of unknown plasmonic nanostructures through simple transmission spectra.The network architecture is established based on simulated data to achieve accurate identification of both geometric and material parameters.We then demonstrate that the network training by a mixture of simulated and experimental data can result in correct material property recognition.Our work may indicate a simple and intelligent recognition approach to plasmonic nanostructures by spectroscopic techniques.Optical bound states in the continuum（BIC）are found in various dielectric,plasmonic and hybrid photonic systems.The localized BIC modes and quasi-BIC resonances can result in a large near-field enhancement and a high-quality factor with low optical loss.They represent a very promising class of ultrasensitive nanophotonic sensors.In this research work,we discovered the quasi-BIC mode in the silicon-based photonic crystal structure by FDTD simulation,and successfully achieved the modulation of the quasi-BIC mode response by changing the thickness of the etched film,which shows good application prospects in refractive index sensing and other aspects.Subsequently,the symmetry-protected BIC resonant dielectric structure is further investigated,by breaking the asymmetry of the structure in different ways,we verified that the quality factor Q and the asymmetry coefficientαsatisfy Q∝α-2,and achieved a large range of BIC response coverage from the visible to near-infrared wavelengths.Finally,we demonstrate a simple neural network-based inverse design where the structural parameters of a periodic all-dielectric hypersurface supporting the quasi-BIC mode can be directly output by inputting the required quasi-BIC resonance wavelength and response Q values.In summary,this thesis proposes an artificial intelligence algorithm-assisted method for the recognition and design of metasufaces.We successfully achieved rapid recognition of plasmonic nanostructure which demonstrate ultra-high accuracy,and are equally applicable to the real experimental data.In addition,we demonstrate an efficient inverse design method for supporting dielectric supersurfaces in quasi-BIC mode,showing great potential for applications.The results of these two works will contribute to solving the challenges of design and recognition in the field of nanophotonics,enabling rapid design and recognition at low cost and high efficiency,which may promote the rapid diffusion of metasurfaces in practical applications such as sensing,imaging and communication. |
PDF Full Text Request