IntroductionProgress in optical design has always been a function of the computational tools available to the designer.In the last half of the Twentieth Century, progress in optical design has been propelled along by vast increases in the power of computing hardware.This development has led to a complete change in the way optical design is actually done.Modern optical design is intensely computer-dependent but still at a point where intelligent inputs from a human designer are essential to the process of optical design.The software tools available today are the result of over 25 years of research, development and use and can provide cost -effective solutions to increasingly complex problems.They can yield impressive results when used by an experienced optical designer and even, in some cases, the novice designer.1 At the same time, the software tools are by no means fail -safe, and the unknowing user may spend dearly for a solution to a problem or even find no solution at all. This paper is intended to review for the optical designer the basic optimization technology used in software tools for optical design.In terms of artificial intelligence, optimization programs are on the level of dumb beasts; in terms of mathematical sophistication, the best programs, in the hands of knowledgeable designers, are capable of large scale problem solving unequaled in the history of technology.The designer must appreciate what problems the algorithms are capable of solving.Indeed, an understanding of what problems an algorithm does not solve is as important in practice as knowledge of the positive attributes of an algorithm's performance.The discussion opens with a statement of the optical design optimization problem. Some preliminary background on the mathematical approach taken is given before the methods themselves are given. This is done in the two main sections: unconstrained and constrained optimization. Before summarizing, some final comments will be offered on the prospects for global optimization and on directions for future optimization research.
Statement of problemThe optical design optimization problem is stated as follows:Given an initial optical system with a specified set of variable parameters, find a new set of parameters which improve an optical performance function subject to physical and optical constraints on the optical system. This is a fairly general statement but not unrealistic in terms of what is encountered in practical design. The problems found in optical design have certain characteristics which direct the choice of the algorithms of interest.Cogent points concerning the elements of the problem statement above are:All methods in use today optimize the performance of an optical system by using an algorithm which minimizes an aberration function expressed as a function of system constructional parameters, the problem variables. It is usually necessary to oversample the aberration function, i.e. have more data points than there are variables to get good correlation between the optical system performa...