2021MNRAS.504.1801K

Previous work has argued that atomic gas mass estimates of galaxies from 21-cm H I emission are systematically low due to a cold opaque atomic gas component. If true, this opaque component necessitates a ∼35 per cent correction factor relative to the mass from assuming optically thin H I emission. These mass corrections are based on fitting H I spectra with a single opaque component model that produces a distinct 'top-hat' shaped line profile. Here, we investigate this issue using deep, high spectral resolution H I VLA observations of M31 and M33 to test if these top-hat profiles are instead superpositions of multiple H I components along the line of sight. We fit both models and find that >80 per cent of the spectra strongly prefer a multicomponent Gaussian model while <2 per cent prefer the single opacity-corrected component model. This strong preference for multiple components argues against previous findings of lines of sight dominated by only cold H I. Our findings are enabled by the improved spectral resolution (0.42 km s^–1), whereas coarser spectral resolution blends multiple components together. We also show that the inferred opaque atomic ISM mass strongly depends on the goodness-of-fit definition and is highly uncertain when the inferred spin temperature has a large uncertainty. Finally, we find that the relation of the H I surface density with the dust surface density and extinction has significantly more scatter when the inferred H I opacity correction is applied. These variations are difficult to explain without additionally requiring large variations in the dust properties. Based on these findings, we suggest that the opaque H I mass is best constrained by H I absorption studies.