Research On Design Method For Broadband Metalens Applications

With the development of science and technology,the requirements for information processing and computing speed are getting higher and higher,which makes integrated and miniaturized optical elements become a development trend.However,optical elements based on natural materials and metamaterials have problems such as large scale and low efficiency.In recent years,metasurfaces composed of artificial microstructure units can precisely control the outgoing beam by designing the local interaction between the incident light and the structural units,and freely design the optical wavefront with sub-wavelength resolution,which provides an unparalleled platform for the integration and miniaturization of optical elements.Among them,metalens,as an important research direction in the field of metasurfaces,has received extensive attention from researchers.Based on the related theory of metasurface optics,this paper studies broadband metalens.Firstly,broadband metalenses operating in the telecom region and the short-wave near-infrared range are designed respectively,demonstrating the great potential of metalens in replacing traditional optical components and improving the performance of optical devices.Then,to solve the chromatic aberration problem of broadband metalenses,a broadband achromatic metalens operating in the mid-wave infrared is designed.Finally,aiming at the problem that the establishment of the broadband achromatic metalens structural unit database requires a lot of computing resources and computing time,a broadband achromatic metalens operating in the visible light range is quickly constructed based on deep neural networks.The specific research work is as follows:(1)For most photonic crystal fibers,high coupling efficiency cannot be obtained because the numerical aperture is too low,and the transmitted light is easy to diverge.In this paper,two novel photonic crystal fiber devices integrating broadband metalens are designed,namely,photonic crystal fiber diverging metalens and photonic crystal fiber focusing metalens with opposite functions.Firstly,the basic knowledge and working principle of photonic crystal fiber,as well as the coupling method of traditional optical fiber,are introduced.Second,two novel photonic crystal fiber devices integrating broadband metalens are designed.Finally,the performances of the designed photonic crystal fiber metalenses are characterized by numerical simulation.By shaping the optical wavefront,the designed device can greatly increase the numerical aperture of the photonic crystal fiber and achieve ideal coupling efficiency.In addition,the photonic crystal fiber focusing metalens can also effectively improve the focusing ability of the photonic crystal fiber to the transmitted light.(2)In recent years,with the miniaturization of the CCD,the fill factor of the pixel has not been significantly improved,which in turn affects the utilization rate of the CCD for incident light.To this end,two broadband metalenses operating in the short-wave near-infrared and insensitive to the polarization of incident light are designed in this paper.The designed metalenses can focus the incident light into an area the size of a CCD pixel to improve the fill factor of the CCD pixel.Firstly,the types and working principles of CCDs are introduced,as well as the applications of short-wave near-infrared CCDs.Secondly,according to the typical pixel size and pixel pitch size of short-wave near-infrared CCD,two metalenses are designed to be compatible with it.Then,the performances of the designed metalenses and the fabricated metalens sample were characterized based on numerical simulations and experimental tests,respectively.The simulated and experimental results show that the designed broadband metalenses can effectively improve the fill factor of short-wave nearinfrared CCD pixels.(3)Due to the high phase dispersion of the structural unit,the broadband metalens has inevitable chromatic aberration,which seriously affects the imaging quality of the optical system.Firstly,the causes of chromatic aberration,traditional achromatic methods,and the research status of achromatic metalens are studied.Second,a database containing a large number of structural elements was created to provide the design parameters required to construct broadband achromatic metalenses.Then,a mid-wave infrared broadband achromatic metalens that is insensitive to the polarization of incident light is designed.The simulated results show that the designed metalens can achieve near-diffraction-limited focusing within the designed bandwidth.Finally,a broadband achromatic metalens array is simulated,and it is proved that the designed metalens can increase the fill factor of the pixels of the mid-wave infrared focal plane array,while effectively suppressing the optical crosstalk between adjacent pixels.(4)The design of each broadband achromatic metalens is inseparable from a database containing enough structural units to provide the design parameters required for achromatic,such as group delay,effective refractive index,and/or phase shift.For each structural unit in the database,numerical simulation can provide accurate results,but it takes a lot of time and computational resources.Aiming at this problem,by introducing deep neural networks,a database of structural units can be constructed quickly and accurately.Firstly,the relevant theories about deep learning are introduced.Secondly,the achromatic principle of metalens based on geometric phase is discussed.Then,the training samples of the structural units are established by the finite-difference time-domain method and trained based on deep neural networks.Through the designed networks,a large number of building blocks can be rapidly obtained,and the group delay and average polarization conversion efficiency of these building blocks can be accurately predicted.Finally,broadband achromatic metalenses in the visible range are constructed based on the training samples and the structural units predicted by the deep neural networks,respectively.The simulation results show that the metalens constructed based on the structural units obtained by the deep neural network can eliminate the chromatic aberration in the visible light range,and at the same time obtain higher average focusing efficiency than those constructed by the training samples.