Effects of molecular contamination of IR near-continuum bandpasses on measurements of M-type Mira diameters

We study the contamination of six near-continuum bandpasses (0.82, 0.92 and 1.04 mum, H, K, L*) that are commonly used for interferometric measurement of diameters of M-type Mira variables. Centre-to-limb variations (CLVs) of intensity predicted by Mira models of Bessell, Scholz & Wood and Hofmann, Scholz & Wood show slight to substantial effects of molecular absorption by H2O (H, K, L*), OH (H), CO (H), TiO (0.82, 0.92, 1.04), VO (0.82) in different models which may lead to spurious observed 'continuum diameters'. In particular, two-component-type CLVs in H, K and L*, composed of the continuum light of the inner disc and the light emitted by H2O molecules in the outer atmosphere may affect visibilities in such a way that UD fits result not only in strongly wavelength-dependent but also, in case of single-baseline observations, in baseline-dependent diameters. Similar effects occur when sub-intervals of the H, K and L bandpasses are considered. Continuum contamination depends on parameters of the Mira models and may vary substantially with variability phase and cycle. Observational evidence supporting model predictions is assembled, and strategies of deriving physically meaningful Mira diameters are outlined. The period-radius relation and consequences for the pulsation mode of Miras have to be rediscussed.