A team of researchers has developed a new framework for interpreting exoplanet observations that incorporates planetary evolution over time, addressing a fundamental limitation of existing static retrieval models.
The work, submitted to the Astrophysical Journal and led by Harrison Nicholls of the University of Oxford, builds on the PROTEUS planetary evolution framework and introduces a method called evolutionary Bayesian retrieval. Current static retrieval frameworks, which take a single snapshot of an exoplanet’s spectrum and infer its atmospheric composition, suffer from degeneracies: different combinations of planetary parameters can produce the same observed spectrum. By modeling how a planet evolves from its initial magma ocean state to its present-day condition, the new approach constrains interpretations to physically permissible scenarios.
The framework uses asynchronous Bayesian optimization to efficiently dispatch PROTEUS forward models in parallel. Each model simulates coupled interior-atmosphere evolution: the planet starts with a fully molten mantle at 3,200 K, which cools and solidifies from the core-mantle boundary upward, while atmospheric outgassing, escape, and radiative-convective climate evolution are tracked simultaneously.
The retrieval jointly infers parameters that are inaccessible to remote sensing, such as mantle redox conditions, metallic core fraction, and initial volatile inventories of hydrogen, carbon, and sulfur, from spectroscopically observable quantities including photospheric radius, temperature, gravity, and atmospheric molecular abundances.
The team tested the framework on three prototypical exoplanets: a young sub-Neptune (3 Earth masses, 1 Gyr old), an older super-Earth (1.9 Earth masses, 2 Gyr old), and a warm terrestrial planet (1 Earth mass, 100 Myr old), all orbiting an M3-type star modeled on L 98-59.
Results showed that degeneracies remain for some parameter combinations, particularly core fraction versus volatile budget for lower-mass planets. However, for terrestrial-mass exoplanets, strong correlations between observable quantities and underlying parameters lifted these degeneracies, allowing post-formation volatile inventories to be recovered with less than 20 percent error.
The paper positions the framework for incoming data from the James Webb Space Telescope, PLATO, the Nancy Grace Roman Space Telescope, and the Extremely Large Telescope. The authors argue that the adoption of time-evolved models will be essential for correctly interpreting the high-quality observations these facilities will produce, and for connecting present-day exoplanet properties to their formation and evolutionary histories.
Images: (none, research paper)
Sources: arXiv:2607.25845 (Nicholls et al., submitted to ApJ, Jul 28, 2026)