Atmospheric regions below a refractive boundary are hidden in limb observations. Refraction thus creates a gray continuum in the planet's transmission spectrum which can hide spectral features associated with sources of atmospheric opacity. We combine refractive theory with recent analytical advances describing the effects of surfaces and clouds on transmission spectra, to express the location of this boundary in atmospheric opacity space, both for atomic and molecular extinction, as well as collision-induced absorption. This allows one to quickly estimate how refraction affects spectral features in well-mixed atmospheres. We show that differences in the geometry of limb observations between solar system planets and exoplanets leads to different locations of this boundary, and that more than four extra scale heights of atmosphere are hidden in exoplanet transits compared to solar system observations of cold gas giants. We explore how the location of this refractive boundary in exoplanet transits changes in a well-mixed isothermal atmosphere with its temperature and composition, the spectral type of the planet's host star, and the size of the planet. We demonstrate that five extra scale heights of atmosphere are hidden in a terrestrial planet with a CO$_2$ atmosphere compared to a helium atmosphere, resulting in a flatter spectrum than from its smaller scale height alone. We provide results for a few exoplanets, notably those in the TRAPPIST-1 system, to help the scientific community estimate the impact of refraction on the size of spectral features without radiative transfer calculations, and thus help refine planned James Web Space Telescope observations.

Detection of the signature of Rayleigh scattering in the transmission spectrum of an exoplanet is increasingly becoming the target of observational campaigns because the spectral slope of the Rayleigh continuum enables one to determine the scaleheight of its atmosphere in the absence of hazes. However, this is only true when one ignores the refractive effects of the exoplanet's atmosphere. I illustrate with a suite of simple isothermal clear Jovian H2-He atmosphere models with various abundances of water that refraction can decrease significantly the spectral slope of the Rayleigh continuum and that it becomes flat in the infrared. This mimics a surface, or an optically thick cloud deck, at much smaller pressures than one can probe in the non-refractive case. The relative impact of refraction on an exoplanet's transmission spectrum decreases with atmospheric temperatures and increases with stellar temperature. Refraction is quite important from a retrieval's perspective for Jovian-like planets even at the highest atmospheric temperatures (1200 K) considered in this paper, and for all stellar spectral types. Indeed, refraction breaks in large part the retrieval degeneracy between abundances of chemical species and the planet's radius because the size of spectral features increases significantly with abundances, in stark contrast with the non-refractive case which simply shifts them to a larger or smaller effective radius. Abundances inferred assuming the atmosphere is cloud-free are lower limits. These results show how important it is to include refraction in retrieval algorithms to interpret transmission spectra of gas giants accurately.

Although the formalism of Lecavelier des Etangs et al. (2008) is extremely useful to understand what shapes transmission spectra of exoplanets, it does not include the effects of a sharp change in flux with altitude generally associated with surfaces and optically thick clouds. Recent advances in understanding the effects of refraction in exoplanet transmission spectra have, however, demonstrated that even clear thick atmospheres have such a sharp change in flux due to a refractive boundary. We derive a more widely applicable analytical formalism by including first-order effects from all these "surfaces" to compute an exoplanet's effective radius, effective atmospheric thickness, and spectral modulation for an atmosphere with a constant scale height. We show that the effective radius cannot be located below these "surfaces" and that our formalism matches Lecavelier des Etangs et al. (2008)'s in the case of a clear atmosphere. Our formalism explains why clouds and refraction reduce the contrast of spectral features, and why refraction decreases the Rayleigh scattering slope as wavelength increases, but also shows that these are common effects of all "surfaces". We introduce the concept of a "surface" cross section, the minimum mean cross section that can be observed, as an index to characterize the location of "surfaces" and provide a simple method to estimate their effects on the spectral modulation of homogeneous atmospheres. We finally devise a numerical recipe that extends our formalism to atmospheres with a non-constant scale height and arbitrary sources of opacity, a potentially necessary step to interpret observations.