Narrow-band guided-mode resonance filters: theory and experiments

Thesis (Ph. D.)--University of Rochester. Institute of Optics, 2003.For more than ten years now, there has been much interest in developing guided-mode resonance (GMR) filters for optical narrow-band filtering applications. Here, analytic methods for simultaneous control over a filter's line shape symmetry, side-band levels, pass-band width, and pass-band shape are studied. Symmetry and side-band levels are managed by an improved version of the so-called thin-jilm method of Wang and Magnusson [Opt. Lett. 19, 919-921 (1994)]. Inadequacies of the original method are related to an unjustified long-wavelength approximation in calculating the effective-index of a grating. Improved results are obtained by using a more accurate effective-index model or rigorous techniques with the thin-film method. Pole-zero analysis is used to derive symmetry conditions and approximate expressions for the side-band levels of a GMR filter. Results of the pole-zero analysis provide a convenient means of evaluating filter performance and reinforce the principles of the thin-film method. Additional control over pass-band width is achieved by incorporating techniques for adjusting the loss of the resonant mode into the thin-film method. Several factors that limit the minimum obtainable pass-band width are discussed. Two methods that provide additional control over pass-band shape are presented. The first involves stacking GMR filters on top of one another such that reflections from each filter interfere to produce a more rectangular pass-band, while the second is to design a filter with two closely spaced resonances that overlap. Filters based on the second technique are designed using pole-zero analysis conditions a maximally-flat pass-band. Experimental results are presented for two GMR filters that demonstrate simultaneous control over symmetry, side-band levels, and pass-band width. Both filters exhibit line shapes that are fairly symmetric on a logarithmic scale and centered about λ = 1550 nm. The first has a spectral width of 1.87 nm and -20 dB side-bands, while the second has a spectral width of 0.43 nm and -25 dB side-bands. Differences between experimental and expected values are attributed to scattering losses