Problem A.2 - Design For Production

Modern multilayer design tools are quite powerful and allow to design very sophisticated coatings featuring various spectral, phase, color and other properties. Unfortunately quite often a nice theoretical design with very promising optical properties does not provide specified characteristics after production. It happens due to a number of reasons, such as thickness deposition errors, deviations of refractive indices from their nominal values, and many other perturbation factors. In this challenge we are trying to simulate multilayer coating production by “virtual deposition” processes. These processes introduce errors to the design thicknesses (Process A1 and A2) and drifts to some layer material refractive indices and extinction coefficients (Process A2). A “virtual spectrophotometer” can be used to obtain a feedback on the deposition of provided design samples.

The main goal of this challenge would be to try to discover ways to compensate for these perturbation factors. For the sake of the challenge induced errors are reproducible from run to run, therefore the final score of any design is not dependent on the run number, and it is very different from the reality. Thus we assume that our virtual deposition is super-stable, and introduces the same errors during the repeated productions of the coating. The levels of errors is not known, even more, they could be different for different layers. This challenge is a simplified model of the real design-production loop, when a coating or a test sample is produced, and its spectral characteristics are measured with a spectrophotometer. The optical engineer uses these characteristics to try to compensate for discovered deposition problems by adjusting the design, or by designing a different, more robust one. A winning design is the design demonstrating the best score after the “virtual deposition” process, i.e., the design with best compensation of these initially unknown perturbation factors.