Neural network-assisted design of GSST-based achromatic metalens with continuously variable focal heights

Abstract

Recently developed achromatic metalens show great potential to replace conventional lenses. However, current long-wavelength infrared (LWIR) achromatic metalens are limited to fixed functions, severely hampers their integration and application. In order to address this limitation, Ge2Sb2Se4Te1 (GSST) and Silicon (Si) are used as materials for continuously variable focus height broadband achromatic metalens (CVFHBAM) at LWIR. Specifically, the multipole decomposition is performed within the operating band (8–12μm) to avoid undesired resonances and ensure the high degree of linearity of the meta-atoms. A tandem neural network is utilized for parameter optimization. Numerical results show that the CVFHBAM is well-corrected for chromatic aberration within the operating wavelength. By varying the crystalline fraction of GSST, focal length can continuously varied up to five times the center wavelength while maintaining achromatic focus, and the average numerical aperture (NA) is varied from 0.345 to 0.454. In addition, by calculating the Strehl ratio values, the transverse dimensions of the corresponding focal spots almost reach sub-diffractive focusing. The average focusing efficiencies are 57.74%, 64.32%, 68.94% and 71.48% respectively, for the four selected crystalline states. The results show that CVFHBAM has a promising application in developing high-resolution LWIR imaging and spectroscopy systems.