An Oxidation Gradient Straddling the Small Planet Radius Valley

We use the open-source numerical atmospheric escape code IsoFATE¹¹1IsoFATE source code: https://github.com/cjcollin37/IsoFATE (Cherubim et al., 2024). A key difference between the previous study and the present is that we added C and O so that the model computes molecular diffusion and escape of H (protium), He (helium), D (deuterium), O (oxygen), and C (carbon). IsoFATE models energy-limited and radiation/recombination-limited XUV-driven photoevaporation, core-powered mass loss, and molecular diffusion to compute variable escape rates for individual species, allowing for mass fractionation of an escaping atmosphere. We model stellar flux evolution as in Cherubim et al. (2024). We assume all escaping species are atomic, which represents an upper limit for photolysis efficiency in the upper atmosphere. This assumption is discussed in Section 5.

IsoFATE simulates atmospheric escape by numerically integrating several differential equations of the general form

After setting the total number of species to 3, expanding out sums and solving for $\Phi_{3}$, we arrive at an expression for a minor species (species 3) escape flux:

We model magma ocean–atmosphere volatile exchange and equilibrium chemistry with a newly developed Python tool kit for computing the equilibrium conditions at the melt-atmosphere interface called Atmodeller (Bower et al., 2025). Given a set of planetary parameters (e.g., surface temperature, planetary mass, radius, mantle melt fraction) and an initial volatile budget, Atmodeller uses experimentally calibrated solubility laws, together with free energy data for gas species (McBride et al., 2002), to determine how volatiles partition between the atmosphere and interior of the planet, ignoring dissolution of volatiles into possible Fe-Ni core phases. We coupled IsoFATE elemental abundances to Atmodeller and achieve mass conservation between the two modules in an open system where mass is lost to space and exchanged between the planetary interior and atmosphere. We include H₂, He, H₂O, O₂, CO₂, CO, CH₄, and deuterium (D), which is treated as chemically equivalent to H, in our model. We specify the mass of the solar composition atmosphere as an initial condition for each planetary simulation, which sets the abundance of each chemical element in our model (H, D, He, C, and O). Then, for a given temperature and mass of H-He-C-O, the partial pressures of volatile gas species at chemical equilibrium are computed following the ideal gas law. These species are allowed to partition between the atmosphere and molten mantle according to solubility laws that are experimentally calibrated for a basaltic composition, since these are most available in the Earth science literature. This yields the converged value of $N_{i}$, elemental species moles, at any given time step in the simulation. Oxygen fugacity, $f$O₂, equivalent to the partial pressure of the dioxygen species in our model, is also calculated at each step by Atmodeller using the same procedure, that is, based on the pressure, temperature and solubility-modulated elemental composition of the gas phase.

At each time step in our coupled model, we compute the mantle melt fraction as a function of planetary mass and surface temperature following Wordsworth et al. (2018). The mantle melt fraction calculation uses a second-order Birch-Murnaghan equation of state to determine the interior structure for an Earth-like planet with a silicate mantle, iron core, and core mass fraction of 0.3, and assumes dry adiabatic convection. We calculate the local mantle melt fraction in the planetary interior as

Our model assumes an upper limit on magma ocean degassing and ingassing efficiency of volatiles because Atmodeller instantaneously re-equilibrates the interior and atmospheric reservoirs when called. This assumption has been often made since the earliest magma ocean models, e.g. Elkins-Tanton (2008), who estimate a complete magma ocean circulation time of 1-3 weeks for terrestrial magma oceans. In the least efficient case, Salvador & Samuel (2023) determine that 0.1 terrestrial oceans of water could take $\approx$ 45,000 yr to degas from a fully molten Earth-mass planet. This timescale is of the same order as a time step in our simulations, i.e., 50,000 yr. For reference, an Earth-mass planet with a 1% solar-composition atmosphere by mass contains $\approx 0.3$ Earth oceans of atomic oxygen, representing an upper limit on how much water can form. Atmodeller is called every 100 time steps, so all supersaturated water would easily degas for such a planet between each call.

We ran a suite of interior/atmosphere evolution models via Monte Carlo simulations over a broad parameter space. For each trial, 500,000 samples were randomly drawn from log uniform grids of initial planet mass $M_{\mathrm{p}}$ between 1 - 20 M_⊕, initial atmospheric mass fraction $f_{\mathrm{atm}}$ between 0.1 - 30%, and orbital period $P$ between 1 - 300 days. The chosen upper limit for $M_{\mathrm{p}}$ is motivated by the predicted threshold for runaway gas accretion and the chosen range of $f_{\mathrm{atm}}$ was motivated by feasible values for our planets calculated with gas accretion models (Ginzburg et al., 2016). Atmospheric mass loss was initiated at 1 Myr. Orbital migration effects were ignored so $P$ was held constant. Model simulations were halted at 5 Gyr—representing the average planetary age in the Milky Way—for the main results, and 10 Gyr in some cases discussed in Section 3. We assume proto-solar relative abundances of H, He, D, O, and C in a planetary atmosphere enveloping an Earth-like rocky core to simulate nebular gas capture (Lodders, 2003), though in reality these values are expected to deviate for different systems. As such, we employ a “dry start” model, ignoring C, H, and O sourced from the planetary interior and sources of volatiles that enhance planetary metallicity such as cometary delivery and planetary formation in ice-rich environments beyond snowlines. Hence, our model does not produce water worlds with water mass fractions on the order of several percent. We simulate escape with XUV-driven photoevaporation alone and the combination of photoevaporation and core-powered mass loss. We ignore scenarios in which core-powered mass loss operates alone, as we expect photoevaporation to dominate in our parameter space of interest (Owen & Schlichting, 2023; Tang et al., 2024). We performed simulations for planets around G, K, and M stars and we focus on the results for planets around M stars given that they are most amenable to atmospheric characterization.

Our Monte Carlo simulations reveal an atmospheric oxidation gradient that increases toward shorter orbital periods and smaller planet radii (Figures 1 and 2). Figures 1a and 1b show interpolated species molar concentrations as a function of orbital period vs. planet radius and scale height, $H=k_{\mathrm{B}}T_{\mathrm{eq}}/\mu g$, where $k_{\mathrm{B}}$ is the Boltzmann constant, $T_{\mathrm{eq}}$ is the planet equilibrium temperature, $\mu$ is the average atmospheric atomic mass, and $g$ is the gravitational field strength. Figures 1c and 1d show the density of simulated planets with atmospheres dominated by each indicated species, except for H₂O, for which the threshold is 5%.

Five broad families of atmospheres with varying oxidation states emerge in the planet radius-orbital period parameter space that determine the dominant atmospheric species: H₂ worlds, He worlds, CO worlds, CO₂ worlds, and O₂ worlds. We exclude H₂O worlds from this list as our simulations produce few planets ($<1\%$) with water-dominated atmospheres, as discussed in Section 5. H₂-dominated worlds represent the most reduced planets, having retained H/He-dominated primordial atmospheres. He worlds have typically lost the bulk of their H, making them depleted in H₂ and typically enriched in CO and CO₂, or, less commonly, CH₄ if they retain sufficient H₂. CO worlds are more oxidized than He worlds. They are H₂ poor, CO₂ rich, and typically have short orbital periods, existing within and just above the radius valley. CO₂ worlds are even more oxidized. Their second most abundant atmospheric species tends to be O₂ and they have smaller radii than the He and CO worlds, situated between the He worlds and O₂ worlds. Finally, O₂ worlds represent the most oxidized planets, possessing O₂-dominated atmospheres. They exist within and below the radius valley, have the smallest radii, and are on close-in orbits. H₂O-rich worlds are rarely H₂O-dominated in our simulations and span a wide parameter space that largely overlaps with the other families, particularly O₂ worlds.

It is important to note that overlapping families in Figure 1 do not necessarily indicate the prevalence of individual planets belonging to multiple families. For example, the CO- and O₂ world groups overlap substantially in panels c and d, but atmospheres cannot possess both CO and O₂ in high abundances. O₂-dominated atmospheres are typically accompanied by CO₂ as the second most abundant gas, and vice versa. The CO and O₂ world groups overlap in $P-R_{\mathrm{p}}$ space because CO-dominated worlds can exist where O₂-dominated worlds can independently exist. Interestingly, CO worlds must start with much smaller atmospheres ($f_{\mathrm{atm,0}}\approx 0.005$) compared with O₂ worlds ($f_{\mathrm{atm,0}}\approx 0.02$). Both end up with similar atmospheric masses after 5 Gyr, and since CO and O₂ have similar molecular masses, their radii are similar for a given planetary mass, resulting in the observed overlap in $P-R_{\mathrm{p}}$ space. This implies that we may infer formation conditions based on a planet’s atmospheric oxidation state, mass, and radius. Potential photochemistry effects are discussed in Section 5.3.

Figure 2 a and b show interpolated normalized oxidation power, $P_{\mathrm{ox}}$, with overplotted contours of the same families presented in Figure 1. Oxidation power, $P_{\mathrm{ox}}$, is determined following Wordsworth et al. (2018). Each elemental species is assigned an oxidation potential, $p_{i}$, which essentially reflects the number of electrons available to exchange in a reduction-oxidation reaction: $p_{\mathrm{O}}=+2$, $p_{\mathrm{H}}=-1$, and $p_{\mathrm{C}}=-4$. For each simulated planet, $p_{\mathrm{i}}$ is summed for all species to get the total oxidation power, $P_{\mathrm{ox}}=\sum_{i}n_{i}p_{i}$, where $n_{i}$ is the total number of atoms of species $i$. Figure 2 a and b show the interpolated $P_{\mathrm{ox}}$ values normalized between -1 and 1. Only planets in the red parameter space have sufficient O to be net oxidized and hence exist within the O₂ world family.

Figure 2 c and d show the O₂ fugacity, $f$O₂, for the same simulated planets plotted in panels a and b. The solid red contours show $f$O₂ values in bar and the dashed green contours indicate the logarithmic ratio between the $f$O₂ (in bar) and that defined by the iron-wüstite (IW) mineralogical buffer: $\Delta$IW = $log_{10}f$O${}_{2}-log_{10}$IW (Hirschmann, 2021). Planets with small radii and lower equilibrium temperatures have greater $\Delta$IW values than small, hotter planets on closer-in orbits. This is because the $f$O₂ in equilibrium with the iron-wüstite buffer increases with temperature. Therefore, at constant $\Delta$IW, higher temperatures result in higher $f$O₂ (in bar). We find a population of planets with lower surface temperatures (mean $\approx 1840$ K) that are more oxidized, relative to IW (above $\Delta$IW = +8), than are those with higher surface temperatures (mean $\approx 3000$ K) and higher absolute $f$O₂ ($>$100 bar, but between IW and $\Delta$IW = +8). For reference, the present-day Earth’s mantle has $\Delta$IW = +3.5 at 1650 K (Frost & McCammon, 2008).

The atmospheric composition evolution of six archetypical worlds orbiting a typical M1 type star from each oxidation family is shown in Figure 3. The archetypes can be categorized into three general groups illustrated in Figure 3: planets with atmospheres in which the dominant gas is H₂ (as for the CH₄ world and the H₂ worlds, not shown), O₂/CO₂ (as for the O₂, CO₂, H₂O, and He worlds), or CO/CO₂ (as for the CO world). The second most abundant species in O₂- and CO world atmospheres is CO₂ and the second most abundant species in CO₂ atmospheres is O₂. H₂O can be abundant or depleted across all groups, but tends to be retained at least at the ppm level across groups. No planets in our simulations form CH₄-dominated atmospheres, although some atmospheres can reach CH₄ molar concentrations up to several percent. The extended 10 Gyr simulations reveal that He worlds are often transient phenomena (discussed more in Section 3.2) and tend to be rich in oxidized species such as O₂, CO, and CO₂. Figure 3 illustrates that planet atmospheric composition can change drastically within the first 5 Gyr of evolution, sometimes earlier. The archetypes show that planets are highly dynamic objects with atmospheres and interiors that change composition significantly over Gyr timescales.

Figure 3 also gives a sense of the typical magma ocean lifetimes for planets in our simulations. Most simulated planets maintain non-zero mantle melt fractions throughout the typical 5 Gyr simulation time, although only $\approx$ 50% of CO₂ worlds do as a result of their low atmospheric masses and wide incident stellar flux range. Roughly one third of He worlds have solid mantles by 5 Gyr, as they occupy a similar stellar flux range but tend to have larger atmospheric masses than CO₂ worlds, leading to greater surface temperatures. Planets with non-H₂-dominated atmospheres with $P\leq 4$ days (semi-major axis $\leq$ 0.04 AU, $T_{\mathrm{eq}}\geq 620$ K) tend to have mantle melt fractions $\approx$ 20 - 80%, while those with $P\geq 50$ days (semi-major axis $\geq$ 0.21 AU, $T_{\mathrm{eq}}\leq 267$ K) tend to have solid mantles. Planets in between have melt fractions between 0 - 20%.

The most oxidized planets in our simulations reside below the radius valley and have thin, high mean molecular weight atmospheres largely supplied by volatiles that outgas from the mantle as it crystallizes. Such planets are visible in Figure 4 and were not produced in our initial IsoFATE simulations, which did not include coupled atmosphere-interior chemistry (Cherubim et al., 2024). The dominant physical mechanism giving rise to such planets is water dissolution in the liquid mantle. Water is highly soluble in silicate liquids (Dixon et al., 1995; Sossi et al., 2023), particularly compared to the other volatile species considered in our system, and the vast majority ($>99~{}\%$) of its mass is sequestered in the mantle early in a planet’s evolution. This has the effect of modulating the rate of atmospheric escape. When water is dissolved in the mantle, it is shielded from escape, thereby lowering the escape rate because the planetary radius decreases. This effect is magnified for larger atmospheres because they produce higher surface temperatures which increases the volume of the magma ocean. As the primordial atmosphere escapes, water gradually outgasses, is photolyzed, and escapes as H and O. Because H escapes more readily than O, this process must result in oxidation of the atmosphere via Rayleigh distillation, given that the budgets of all volatile elements are finite.

The water shielding mechanism allows planets to retain atmospheres that would otherwise lose them entirely and it allows H₂ to be retained on longer timescales (several Gyr). The mechanism and its impact on three archetypical worlds is shown in Figure 5. The dashed lines represent simulations in which water solubility in the mantle was excluded and the solid lines represent simulations with solubilities included for all species. The blue shaded regions highlight the amount of water retained as a direct result of the water shielding mechanism enabled by water dissolution in the mantle. For the O₂ world, without water shielding, the entire atmosphere would be stripped within 1 Gyr. In the case of the CO world, water shielding is shown to be a crucial mechanism for atmospheric oxidation resulting in CO as the dominant species. In all cases, substantially more H₂ is retained at least up to 10 Gyr.

Sub-Neptunes with fractionated atmospheres enriched in He (Hu et al., 2015; Malsky & Rogers, 2020; Malsky et al., 2023; Cherubim et al., 2024) and D (Gu & Chen, 2023; Cherubim et al., 2024) have been robustly predicted by atmospheric escape models. Our IsoFATE-Atmodeller coupled model reproduces the same prominent population of He worlds along the upper edge of the radius valley as previously predicted with IsoFATE in Cherubim et al. (2024). The He worlds can be seen in Figures 1, 2, and 4. The strongly D-enriched planet population predicted by Cherubim et al. (2024) is also reproduced, but the correlation between D/H and atmospheric surface pressure is weaker (Figure 6). This is due to the water shielding mechanism described in Section 3.1. Water is the dominant H carrier for a planet that has lost the bulk of its primordial (i.e., H-, He-rich) atmosphere and is highly soluble in the magma ocean. This serves as a vast H reservoir that provides a steady supply of H to the atmosphere, thereby diluting the elevated D/H ratio in the atmosphere. As a result, thinner atmospheres are allowed to have lower D/H ratios than previously predicted in IsoFATE simulations without magma ocean coupling. Our simulations suggest that D/H may still be used to place an upper limit on atmospheric mass and surface pressure. While most of our D-enriched planets peak near Earth’s value of $\approx$ 10 $\times$ solar, many planets exceed Venus-like values and can even be depleted entirely of H in place of D. Planets with enhanced D/H tend to have $M_{p}<$ 5 M_⊕, averaging $\approx 2$ M_⊕; orbital periods between 1 - 150 days, averaging 20-30 days; and initial $f_{\mathrm{atm}}\leq 1\%$.

While the He world population is robustly predicted and dominates much of the parameter space, Figure 3 demonstrates that they are actually transient states that typically exist on the order of hundreds of Myr to several Gyr. The archetypes in Figure 3 were selected based on their atmospheric compositions at 5 Gyr, which we take to be a representative age for an average exoplanet around a main sequence star in the Milky Way. However, when extending the simulations to 10 Gyr, we see that He is not stable on longer timescales and is gradually lost. If the He inventory manages to survive the bulk EUV emission (within 500 Myr for M stars in our model), a He-dominated atmosphere can remain stable for several Gyr. However, even as EUV emission rapidly declines after the stellar saturation phase, the energy received by the planet is sufficient to drive off He. Unlike H₂O, which can be sequestered in the mantle during the magma ocean phase, and gradually released as the mantle crystallizes, He is inert and is relatively insoluble in magma, leaving it susceptible to escape (Jambon et al., 1986). The He world survival timescale is therefore sensitive to assumptions about planetary mass, surface temperature, and stellar evolution. For simulated planets around M stars whose atmospheres become He-dominated at some point in their lifetimes, the mean survival timescale is 900 Myr and the median is 380 Myr.

An additional factor that could contribute to prolonging a He world’s lifetime is radiogenic He produced by uranium and thorium decay in the mantle. Extrapolating from the estimated He outgassing rate from Earth’s surface today, we expect $\approx 1\times 10^{18}-1\times 10^{21}$ mol/Gyr for rocky planets in the 1 - 10 M_⊕ range (Krasnopolsky et al., 1994). For common He escape rates, e.g. $\approx 1\mathrm{x}10^{23}$ mol/Gyr for the archetypical He world, radiogenic He may be insufficient to replenish escaping He in most cases, however the Earth estimate may represent a lower bound since He outgassing on planets with magma oceans is not expected to be limited by tectonic activity as it is on Earth.

The prominent O₂ world population seen in Figures 1 and 2 comprises $\approx 70\%$ of all simulated planets with non-H₂ secondary atmospheres. This represents an upper limit on atmospheric oxidation and O₂ enrichment because our magma ocean model ignores iron oxidation reactions. That is, if the magma ocean is communicating with the atmosphere throughout a planet’s lifetime, then the presence of Fe dissolved as Fe²⁺ and Fe³⁺ in the mantle of the planet could buffer the O₂ levels in the atmosphere to lower proportions than those shown in Figures 1 and 2. We performed simulations to test the sensitivity of O₂ world occurrence to assumptions about iron abundance and oxidation state in the mantle, following the approach in Schaefer et al. (2016) and Wordsworth et al. (2018). We assumed mantle iron mass fractions between 0.05 - 0.4 and that all iron was initially in a reduced state: Fe²⁺. At each time step of our numerical simulation, any atmospheric O₂ buildup was removed to oxidize Fe²⁺ to Fe³⁺ via 4FeO + O₂ $\rightarrow$ 2Fe₂O₃. This reaction proceeded until all Fe²⁺ was oxidized to Fe³⁺, at which point O₂ was allowed to accumulate in the atmosphere. The O₂ world population is recovered in these simulations and is robust to the mantle iron sink, as seen in Figure 7. O₂ world occurrence decreases with increasing mantle iron mass fraction, as O in the atmosphere is consumed through the oxidation of ferrous iron, but is only marginally affected by mantle iron mass fractions estimated for the rocky solar system bodies, e.g. $\approx 6\%$ for Earth (McDonough & Sun, 1995). That the occurrence of O₂ worlds is largely insensitive to the presence or absence of a mantle iron sink, despite Fe being more abundant than O, revealed that the vast majority of such worlds tend to have smaller and/or more short-lived magma oceans, preventing all of the atmospheric O₂ from being consumed by the reduced iron reservoir in the mantle. In other words, mantle freezing outpaces O₂ production sufficiently to preclude substantial O₂ sequestration via mantle iron oxidation.

A competing factor to atmospheric O₂ buffering due to mantle iron oxidation is enhanced planetary metallicity. Our simulations generate sub-Neptunes that possess C, H, and O in solar abundances and hence are initially “dry,” though this ignores O in the interior, which could combine with the H delivered in the atmosphere to form H₂O (see above). For planets that inherit additional water, for example through cometary delivery or accretion of icy planetesimals during formation, the O inventory can be enhanced by several orders of magnitude. If the O abundance is sufficiently enhanced on wet planets endowed with larger water inventories, O₂ may continue to build up in the atmosphere as mantle iron oxidation saturates, H₂O is photolyzed, and H escapes. Tian (2015) showed that atmospheric O₂ buildup correlates with initial water inventory for Earth analogues around early M stars undergoing atmospheric escape.

The main controls on interior-atmosphere exchange are magma ocean volume and volatile solubility, which are determined primarily by surface temperature and mantle composition respectively. To determine surface temperature, our model assumes a dry adiabat in a troposphere defined at atmospheric pressure greater than 0.2 bar (Robinson & Catling, 2012). This requires an assumption about the adiabatic index, and thus the specific heat capacity of the gas mixture in the atmosphere, which we take to be $R/0.28$ where $R$ is the specific gas constant. For reference, $R/c_{p}=0.28$ for a N₂ atmosphere, 0.29 for an H₂ atmosphere, 0.4 for a He atmosphere, and 0.22 for a CO₂ atmosphere at 300 K. The surface temperature calculation is highly sensitive to the choice of adiabatic index and the only downstream calculation affected by surface temperature in our model is mantle melt fraction.

In order to evaluate the sensitivity of our main results to the surface temperature and mantle melt fraction calculations, we repeated our simulations assuming a fixed mantle melt fraction of 1.0 to represent an end member case. We found that the key result of a planetary oxidation gradient surrounding the small planet radius valley is largely unaffected, although D/H fractionation was significantly muted, rarely exceeding $\approx$ 50 times the solar value, as more H₂O shielding provided a larger reservoir to dilute the D/H enhancement. Our previous results represent the other extreme in which the mantle melt fraction was fixed at 0, i.e. no interior-atmosphere exchange was allowed (Cherubim et al., 2024).

It is possible, especially for high mean molecular weight secondary atmospheres, that convective inhibition near the planetary surface may form a radiative layer which lowers the surface temperature (Selsis et al., 2023; Innes et al., 2023). This would have the effect of reducing the mantle melt fraction and magma ocean volume, reducing water shielding and mantle oxidation. We assumed zero planetary albedo, $A$, for all simulations in our main results, which places an upper limit on equilibrium temperature and its influence on surface temperature. We repeated simulations assuming $A=0.5$ and revealed that, while individual planets have different outcomes due to lower equilibrium and surface temperatures, the $(1-A)^{(1/4)}$ equilibrium temperature dependence is insufficient to impact the planet population in our model.

Our escape rate calculations are affected by uncertain factors including stellar flux history, escape shutoff via molecular line cooling, and the effects of ionization on atomic and molecular diffusion in the escaping wind. We adopt a simple power law parameterization to model the stellar EUV flux evolution which ignores the pre-main sequence, during which excess flux may drive additional escape. We repeated simulations around M stars using an EUV evolution model that includes the impact of stellar rotation on magnetic field evolution which captures behavior in the pre-main sequence phase (Johnstone et al., 2021). Our key results are unaffected. This is due in large part to the radiation/recombination-limited escape calculation which limits the escape rate in spite of elevated flux during the pre-main sequence phase. Hence, the integrated EUV energy that effectively drives escape over 5 Gyr is comparable to that for simulations using the power law parameterization.

Our model ignores ionization of escaping species, except for the role of ionized hydrogen in radiation/recombination-limited escape. Ionized atomic species are known to possess significantly larger collisional cross sections due to Coulomb interactions, which will impact species diffusion in the outflow (Schulik & Booth, 2023). This could potentially suppress D/H fractionation, for which the mass ratio is minimal, but is unlikely to overwrite the general predicted oxidation gradient. Our radiation/recombination-limited escape calculation accounts for dissipation of high energy radiation via recombination of protons and electrons and subsequent emission in Lyman-alpha, and captures some of the physics regarding escape suppression due to line cooling. However, our model ignores other line cooling mechanisms that may occur in the presence of the included molecular species (Nakayama et al., 2022; Yoshida et al., 2024). Misener & Schlichting (2021) propose that the core-powered mass loss mechanism may shut down when the cooling timescale equals the loss timescale, leaving behind a tenuous atmosphere. They report analytical estimates of retained atmospheric mass fractions between $10^{-4}$ and $10^{-8}$ which largely coincide with values produced in our simulations.

Photochemistry may affect our results in predictable ways. First, a dry CO₂ atmosphere would likely be susceptible to a CO-runaway state (e.g. Zahnle et al., 2008; Hu et al., 2020), unless trace gases were available to catalytically oxidize the CO like Cl-chemistry does on Venus for example (McElroy et al., 1973). Escape of O could also be faster on such worlds. CO${}_{2}^{+}$ from photoionization of CO₂ is short-lived in the presence of O generated from photolysis of CO₂, and would produce O${}_{2}^{+}$. Subsequent dissociative recombination of O${}_{2}^{+}$ is the primary driver of O loss on Mars today and ion loss of O${}_{2}^{+}$ could have been faster in the past (Jakosky et al., 2018). Second, on CH₄ worlds, hazes are expected to form as a result of CH₄ photolysis, and their formation is generally controlled by an atmosphere’s C:O ratio. Yet, CH₄ and H₂O are unlikely to co-exist at relatively large abundances. OH generated from water photolysis is a strong oxidant and would likely attack products of CH₄ photolysis before haze formation could occur. Finally, on all planets, photolysis of escaping CO₂ and H₂O would split these molecules into CO and OH respectively, which may suppress hydrodynamic escape rates due to radiative cooling effects (Yoshida et al., 2024).