FIGURE 2. Measured V² and fluxes, compared to the predictions of simple physical models.

Plotted error bars indicate estimated 1 (standard deviation) uncertainties of 3%; as noted above, the uncertainty in the shortest wavelength V² measurement is larger, approximately 10%. The dotted line shows the values computed for a model including continuum emission from warm dust and hot gas, and Br emission from hydrogen. This model reproduces the overall slopes seen in the data as well as the feature at 2.165 m, but deviates from the data between 2.0 and 2.1 m. Hot water vapour contributes significant emission at these wavelengths²⁰, and the solid line shows the predictions of a model that includes H₂O. Because the H₂O emission lies at a radius only slightly larger than that of the hot gaseous continuum, its effect on the visibilities is more subtle than its effect on the measured fluxes. Absorption due to terrestrial atmospheric water vapour affects the data in the shortest-wavelength channel, but not substantially in longer-wavelength channels because, in contrast to the hot water vapour in our model, water vapour at atmospheric temperature and pressure conditions lacks substantial opacity at these wavelengths²⁰. The model including H₂O fits the data well over the entire observed wavelength range, and yields substantially smaller residuals between model and data than the model without water vapour. We note that the inferred sizes of the emission components in our model are smaller than the uniform disk sizes plotted in Fig. 1; this is due primarily to the different characteristic sizes for uniform disk and relatively narrower ring models¹⁵. Additional details of the modelling and fitting procedure can be found in previous work¹⁷.