A preprint's numerical models link the chemistry of a basal magma, the molten material at the envelope base, to sharply different atmospheric predictions for sub-Neptune planets, including changes in upper-atmosphere gases and modeled size. The work is a calculation of assumed scenarios, not a measurement of actual worlds, and its predictions depend on the chemical and physical choices built into the model.
The sharpest contrast appeared in models that varied the starting ratio of silicon to oxygen in the basal magma. In the oxygen-rich Si:O=1:4 case, the modeled atmosphere was rich in water, with a mean molecular weight of 3.88 atomic mass units and a transit radius of 1.89 Earth radii. That radius was 30% below the corresponding Si:O=1:2 case. At the Si:O=1:1 end of the set, the atmosphere was oxygen-depleted and rich in silane, or SiH4.
A different starting mix, a different atmosphere
A second set of models added magnesium, carbon and hydrogen to the silicon-oxygen system. In the Mg:Si:O=1:1:3 case, nearly all the oxygen was predicted to rain out into condensates including silicon dioxide and magnesium silicate. The upper atmosphere was strongly reducing, meaning it favored chemical species formed in oxygen-poor conditions, and methane and silane dominated. Silicon carbide began condensing below a few bars of pressure.
The neighboring Mg:Si:O=1:1:4 case followed a different condensate sequence: magnesium silicate, silicon dioxide, Mg2SiO4 and magnesium oxide. Its modeled atmosphere retained oxygen-bearing gases. Carbon chemistry shifted from methane to carbon monoxide and carbon dioxide near 1 bar, a change that illustrates how a small adjustment in the assumed starting composition can place the model in another atmospheric regime.
The study's conclusion describes this change as an abrupt transition under the chosen model settings, despite smoother changes in the gas composition at the base of the envelope. It places the reported boundary at approximately and . Here, the ratios compare the modeled basal abundances of magnesium, silicon and oxygen; the symbols mean approximately greater than or equal to and approximately less than or equal to. The paper attributes the tipping behavior to changes in the order of condensation, but the reported threshold differs between parts of the manuscript, so it should not be treated as a universal cutoff.
The chemistry changed the modeled size too
The differences were not limited to gas inventories. In one fixed comparison, a pure-hydrogen reference envelope had a transit radius of 2.77 Earth radii. The oxygen-poor and oxygen-rich cases measured 2.51 and 2.01 Earth radii in the models. The condensation cases also developed several deep non-convective regions, separated by detached convective zones, rather than the simpler structure of the pure-hydrogen reference.
Across a modeled sub-Neptune mass range, the condensation-associated structures produced radii at least 10% smaller under one set of fixed inputs: a 4,000 K envelope-base temperature, an envelope hydrogen mass equal to 0.03 planet masses, and an equilibrium temperature of 1,000 K. These curves are useful for comparing the model cases, but they do not describe planets with a common age because the base temperature was held fixed.
The calculations were built into a framework called Rocky Raccoon, which couples the FASTCHEM COND chemical network to an envelope-structure model containing hydrogen, magnesium, silicon, oxygen and carbon species. The chemistry stepped downward in pressure, removed material that had condensed at the preceding level, and stopped at an outer boundary of 10−3 bar. The structure solver assigned the transit radius to the 20 mbar level.
Possible clues in spectra, with important caveats
The team also generated synthetic transmission spectra, which simulate how the atmosphere might filter starlight during a transit. Spectra from different assumed interior compositions were distinguishable in the comparison and included features from water, carbon dioxide, methane and silane. Their predicted appearance changed with the assumed quench pressure, the level where the model stopped treating the chemistry as fully equilibrated. Because these were simulated spectra, they do not show that real observations would recover a planet's basal composition.
Other sweeps showed why the result depends on where and how the atmosphere is examined. At fixed base conditions, changing the planet's equilibrium temperature left the modeled gas mixture at the envelope base unchanged but altered abundances higher up; in oxygen-rich cases, methane and carbon monoxide exchanged dominance near 900 K. Changing the atmospheric mass fraction from 0.003 to 0.15 produced base compositions that differed by orders of magnitude, while the compositions at observable pressures were essentially identical. The higher-mass cases had more reducing compositions at the base.
The largest qualification is chemical. The authors say the Gibbs energies used to calculate equilibrium and the resulting condensate speciation are extrapolated far beyond experimentally verified conditions. In particular, adding solid SiO to the 1:1:3 network was associated with a switch from an oxygen-poor to an oxygen-rich upper atmosphere and with a smaller transit radius, while SiO was the dominant deep silicate condensate in that model. That sensitivity makes the allowed condensate list a major part of the prediction.
The paper calls for improved experiments or molecular-dynamics modeling to test the thermochemical inputs used in these calculations. The synthetic comparisons produced distinguishable molecular features, but they do not validate retrievals from real observations. The results therefore remain conditional on the assumed condensate thermodynamics and model chemistry.
Paper data and sources
Original title: Magnesium silicate condensation in sub-Neptune envelopes: the fundamental link between chemistry, structure, and observables
Authors: William Misener, Peter Gao, Nicole L. Wallack et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text