Application of the global discrete-continuous optimization method with selective variables averaging to design of a fast NIR lens

Abstract

Design of an optical system implies definition of its' parameters including both continuous and discrete variables. The first group is corresponds to such parameters as radii of curvature and axial thicknesses, and the second one consists of the optical materials types. A number of algorithms to optimize both groups of variables simultaneously was developed and implemented in optical design software. However, as the working spectral range expands and requirements to the systems' aperture and performance increase, the efficiency of existing design tools for mixed-variables optimization may appear to be insufficient. On top of this, the standard optimization tools do not provide all the necessary control and customization options Therefore, we consider a custom optimization tool to perform a global search in mixed variables. It is based on the method of global optimization with selective averaging of variables. A positive selectivity coefficient is introduced into a positive decreasing functional kernel. With increase of the coefficient the averaging provides convergence of the target discrete variables to the optimal solution. We apply this principle to develop a custom optimization tool. It is used for optimization of an f/1.8 objective lens working in the NIR range of 0.9-1.8 microns with the field of view of 10 deg. We analyze the optimization process convergence in the continuous and discrete variables space and compare our results with the existing optimization tools.