Oxygen left behind: Atmospheric Enrichment due to Fractionation in Sub-Neptunes using BOREAS Released on August 10, 2026 The package developed for this project is publicly available under a custom license and can be found in https://github.com/ExoInteriors/BOREAS . Marilina Valatsou, Caroline Dorn, Pierlou Marty, James E. Owen Abstract The evolution of exoplanetary atmospheres is strongly influenced by atmospheric escape, particularly for close-in planets. Fractionation during atmospheric loss can preferentially remove lighter elements such as hydrogen, while retaining heavier species like oxygen. In this study, we investigate how and under what conditions hydrodynamic escape and chemical fractionation jointly shape the mass and composition of exoplanet atmospheres, especially for mixed H 2 + H 2 O atmospheres. We develop BOREAS , a self-consistent mass loss model coupling a 1D Parker wind formulation with a mass-dependent fractionation scheme, which we apply across a range of planet masses, radii, equilibrium temperatures, and incident XUV fluxes, allowing us to track hydrogen and oxygen escape rates at different snapshots in time. We find that oxygen is efficiently retained over most of the parameter space. Significant oxygen loss occurs under high incident XUV fluxes, while at intermediate fluxes oxygen loss is largely confined to low-gravity planets. Where oxygen is retained, irradiation is too weak to drive significant escape of hydrogen and thus limiting atmospheric enrichment. By contrast, our model predicts that sub-Neptunes undergo substantial atmospheric enrichment over ∼ \sim 200 Myr when hydrogen escape is efficient and accompanied by partial oxygen entrainment. Notably, our results imply that sub-Neptunes near the radius valley can evolve into water-rich planets, in agreement with GJ 9827 d. Present-day water-rich atmospheres may have originated from water-poor envelopes under some conditions, highlighting the need to include chemical fractionation in evolution models. BOREAS is publicly available. I Introduction The rapid increase in exoplanet detections over the past two decades has dramatically improved our understanding of planetary formation, evolution, and atmospheric dynamics. Among the most striking features of the exoplanet population is the wide diversity of atmospheric compositions, spanning from hydrogen-dominated primordial envelopes to This diversity is widely interpreted—especially for close-in planets—as the outcome of multiple evolutionary pathways, including primordial gas accretion, secondary atmosphere production through volatile outgassing, and atmospheric mass loss. The latter process is driven by intense stellar radiation, and can alter atmospheric composition, surface conditions, and long-term habitability ( 36 ; 44 ) . Volatile loss due to irradiation is also believed to have sculpted the terrestrial planets of our own Solar System, as demonstrated by isotopic and noble-gas signatures for Venus, Earth, and Mars ( 19 ) . Observations of sub-Neptunes suggest a range of possibilities: some appear as steam worlds ( 37 ) , while others may retain hydrogen-dominated envelopes, such as TOI-270 d ( 10 ) and K2-18 b ( 24 ) . Although there is considerable debate over whether sub-Neptunes are predominantly gas dwarfs or water worlds ( 2 ) , recent global chemical equilibrium models indicate that all sub-Neptunes likely form water-poor, with bulk water mass fractions <2 %, and initial water envelope mass fractions below 20 % ( 46 ) . Taken together, these results suggest that fractionated mass loss during a planet’s evolution may drive the atmospheric diversity of sub-Neptunes, since lighter species like hydrogen escape more readily than heavier ones, like oxygen. Here, we explore this hypothesis. Loss of volatiles such as water and hydrogen is crucial in determining whether a planet keeps its atmosphere or becomes a bare rocky world ( 22 ) . Hydrodynamic escape is a primary driver of atmospheric mass loss, particularly in young planetary systems, where high-energy stellar radiation ionizes atmospheric gases and powers outflows that can strip hydrogen-rich envelopes and transform sub-Neptunes into super-Earths over geological timescales ( 20 ; 15 ; 40 ) . The efficiency of this escape depends on stellar irradiation, planetary gravity, and atmospheric thermodynamics, with regimes such as energy- and recombination-limited escape setting mass loss rates ( 26 ; 31 ) . Both photoevaporation and core-powered mass loss shape the radius distribution of super-Earths and sub-Neptunes. Rather than competing, they act sequentially depending on system evolution and energy budget: photoevaporation is driven by stellar ultraviolet and X-ray (XUV) radiation ( 18 ) , while core-powered loss relies on residual cooling luminosity ( 11 ) . Which process dominates depends on the sonic point’s location relative to the XUV absorption depth. Population models reproduce the observed radius valley, highlighting photoevaporation as the primary sculptor, while core-powered loss remains important for weakly bound, highly irradiated planets ( 33 ) , though 21 find neither mechanism universally dominant. Here, we adopt a photoevaporation-dominated framework and do not explicitly model core-powered mass loss, although we can identify such outflows within out modeling framework. In addition to mass loss, fractionation plays a key role in shaping atmospheric composition over time. For example, high-energy stellar radiation can dissociate the water molecules of a steam atmosphere into hydrogen and oxygen atoms, whose escape is governed by diffusive separation and hydrodynamic drag ( 48 ) . Because hydrogen escapes more easily than oxygen, the atmosphere may become enriched in heavier elements—a trend supported by studies of transonic outflows in the Solar System ( 13 ; 49 ) and recent exoplanet models. 7 showed that the degree of fractionation is highly dependent on the escape mechanism, with photoevaporation and core-powered loss producing distinct signatures in deuterium and helium. Preferential hydrogen loss can even cause abiotic O 2 buildup, a possible false biosignature as shown by , though other models suggest oxygen is largely dragged out with hydrogen ( 5 ) . The application of fractionated loss at the population level 6 has already revealed strong oxidation gradients and species-dependent loss signatures, demonstrating that coupled fractionation–escape modeling is feasible at scale. Yet, most models treat mass loss and fractionation separately ( 6 ) , leaving key questions unresolved. Full radiation hydrodynamic models could address this but remain computationally challenging for large populations. Our semi-analytic hydrodynamic framework combines a generalized isothermal Parker-wind approach with self-consistent XUV penetration and recombination-limited physics to enable efficient exploration of a broad parameter space. Unlike simple energy-limited scaling models, where the XUV absorption radius is often assumed to match the planet’s radius, our method calculates the XUV radius; these are also the calculations that allow us to determine the transition from core-powered mass loss to photoevaporation. Additionally, it incorporates ionization-recombination equilibrium where appropriate, and enforces a matching condition for the sound speed and mass-loss rate without requiring full radiation-hydrodynamic simulations. At the same time, it simplifies certain aspects, such as the photo-dissociation efficiency and the detailed energy deposition, compared to full radiation hydrodynamic codes. We explore simulations for a population of planets representative of the observed super-Earth and sub-Neptune population. Lastly, we benchmark our coupled model against existing mass loss rate estimates for well-characterized exoplanet systems (see Section VII.1 ). The detailed description of the photoevaporation framework is provided in Section II , and the fractionation methodology is outlined in Section III , with the results of applying it to super-Earth and sub-Neptune environments in Section VI . II Atmospheric Escape Model We build an XUV-driven photoevaporation model based on the modified Parker wind framework of 33 . This approach accounts for both energy-limited and recombination-limited escape regimes, which depend on the atmospheric density and irradiation conditions. In the energy-limited regime, escape is driven by XUV heating, while in the recombination-limited regime, intense ionization forces hydrogen in a dense flow and radiative recombination becomes the main cooling mechanism. This recombination cooling thermostat sets the outflow to a temperature of approximately 10^{4} K and thus caps the mass loss rate. The regime classification and diagnostics follow 33 . The mass-loss framework is coupled with a stellar XUV model that builds on 1 and 40 . II.1 Stellar Irradiation Model To calculate the received XUV flux F_{\mathrm{XUV}} for a planet, we first estimate the star’s XUV luminosity L_{\mathrm{XUV}} . We assume that stars follow a characteristic ratio L_{\mathrm{XUV}}/L_{\mathrm{bol}} , which we interpolate as a function of stellar mass and age using the evolutionary tracks of 1 and the empirical relations compiled in 40 . Although this ratio varies weakly with stellar mass at fixed age, it captures the essential trend that younger and more massive stars emit higher XUV radiation than older or lower-mass stars. Once L_{\mathrm{XUV}} is determined, the incident flux F_{\mathrm{XUV}} at the planet’s orbit is calculated directly from this luminosity and the orbital distance. To establish the relevant orbital distance, we link the planet’s equilibrium temperature T_{\mathrm{eq}} to the star’s bolometric luminosity L_{\mathrm{bol}} using the standard planetary energy balance relation, which equates the absorbed stellar flux to the planet’s thermal re-radiation. Specifically, we assume a Bond albedo \alpha=0.3 , a re-radiation fraction \beta=0.75 , and an emissivity \epsilon=1 . While these parameters are fixed for the models presented here, they remain configurable inputs to our framework for future analyses. Finally, we adopt a heating efficiency of \eta=0.3 , which represents the fraction of absorbed XUV energy that effectively drives atmospheric escape rather than being radiated away from the planet. This value is consistent with prior photoevaporation models ( 33 ) and is found to be appropriate for the low density planets we consider in this work ( 32 ; 17 ) . Figure 1: Time evolution of stellar XUV luminosity for stars of different masses, ranging from 0.1 to 1.4 M ⊙ . II.2 Hydrodynamic Escape Approach We model atmospheric outflow using a 1-dimensional, spherically symmetric, steady-state Parker wind model following the methodology of 33 . The general structure of the modeled atmosphere used to calculate mass loss is illustrated in Figure 2 . We approximate the radiative region above the radiative-convective boundary R_{\mathrm{rcb}} as isothermal at the planet’s T_{\mathrm{eq}} , extending outward to the XUV absorption radius R_{\mathrm{XUV}} . This treatment follows prior work on weakly luminous irradiated atmospheres ( 34 ; 35 ) . Beyond R_{\mathrm{XUV}} , stellar XUV photons heat the gas to higher temperatures T_{\mathrm{outflow}} , driving a photoevaporative outflow. The outflow remains subsonic from the base of the wind up to the sonic point at R_{s} , beyond which it transitions to supersonic expansion. Within this structure, at smaller radii than R_{\mathrm{XUV}} , we also define the transit radius of the planet R_{\mathrm{transit}} . Finally, the Bondi radius R_{B} associated with a locally isothermal outflow is: R_{B}=\frac{GM_{p}}{2c_{s}^{2}}, (1) where c_{s} is the local sound speed, and M_{p} is the planetary mass. For significant photoevaporation to occur in the sense of 33 , we generally require R_{\mathrm{XUV}}R_{s} , mass loss can still occur; we treat these cases using the modified Parker-wind branch discussed in Appendix A.1 . However, if R_{\mathrm{XUV}}>R_{B} , this scenario would correspond to a core-powered mass loss regime. , we do not model it here. Figure 2: Schematic depiction of the model’s atmosphere, adapted from 33 . For a planet undergoing significant photoevaporation, R_{B} lies outside R_{\mathrm{XUV}} , and R_{s}