Following the simulation in (https://mccullerlab.com/logs/lab/index.php?callRep=11455), I decided to use an optic with a constant thickness and material, instead of the substrate having the same thickness. Thus, the whole optic (coating+substrate+coating) is 2 mm thick. For the thick coating, each layer is 100 um. For the thin coating, each layer is 20 um. Interestingly, the location of the peaks changes even though the entire optic is the same thickness. This suggests the loss angle affects the eigenfrequencies, at least in the COMSOL model.