Thin film narrow band optical filters

Abstract

Narrow band filters are interesting for a number of reasons. Firstly there is the simple reason that we want to be able to make them better with a higher process yield. Secondly there is the reason that their performance is very sensitive to alterations in material properties and so they could be useful test beds for materials and especially for detecting small instabilities with time. Techniques for design of thin film narrow band optical filters are well established. The filters may be considered as arrangements of tuned cavities: the simplest type, the Fabry-Perot, consists of just one cavity; the double half-wave (DHW) has two cavities; the triple half-wave (THW) consists of three cavities; and so on. The narrowest filters are constructed completely from dielectric materials to keep dissipative losses as low as possible. Almost all the all-dielectric filter designs in common use employ layers that have optical thicknesses which are integral multiples of quarter of the peak wavelength. This is because both the theoretical treatment and the control of deposition are rather simpler for quarter-waves than for non-integral thicknesses. From time to time proposals have been made for new narrow band filter designs which involve nonintegral thicknesses but these do not seem to have been taken up to any extent by manufacturers, probably because the additional complication outweighs any advantage which they might have in performance. The theory and techniques of layer thickness control are apparently rather less well organized than design. These are central to the production of coatings and so manufacturers are naturally reluctant to discuss their methods in detail. There has not been a great deal of interest in this aspect by academics and as a result relatively little has been published. In this review we consider briefly a few of the problems which occur in the production of narrow band all-dielectric thin film optical filters. For the reasons mentioned above we shall consider only designs composed entirely of optical thicknesses which are integral multiples of a quarter-wave. Assuming that we have stable thin film materials with accurately predictable losses then we can estimate the best performance which should be achieved with a particular design. Real manufactured filters will always have poorer performance than this because errors in the layers will disturb the optimum matching. The resulting decrease in peak transmittance will be accompanied by a corresponding increase in reflectance which is always a good test of whether low transmittance is due to losses or to manufacturing inaccuracies. This leads us directly to the idea of tolerances. How accurately