The atmospheric vertical structure of Uranus and Neptune from thermochemical models: the impact of model assumptions

The atmospheric vertical structure of Uranus and Neptune from thermochemical models: the impact of model assumptions

This paper investigates the atmospheric vertical structure of Uranus and Neptune using thermochemical equilibrium models, focusing on how different model assumptions affect the inferred composition and cloud structure. The authors use the open-source chemical equilibrium code FastChem to compute the vertical distribution and cloud decks of CH4, NH3, H2S, H2O, and NH4SH across a broad range of parameters.

The study varies three key parameters: atmospheric metallicity (from 1 to 80 times solar), elemental ratios (C/O from 0.1 to 2.0, and S/N from 0.19 to 1.6), and 1-bar temperature (from 66 to 86 K). For the thermal profiles, the authors use dry and moist adiabats following Leconte et al. (2017), with H2O as the reference condensing species. They also consider the possibility of convection inhibition and radiative layer formation when molecular weight gradients stabilize the atmosphere.

Key findings include:

1. Metallicity effects: Increasing atmospheric metallicity proportionally increases mixing ratios and cloud thickness. Within the 1 to 80 solar range, abundances of H2O, CH4, and H2S can increase by a factor of 50, while NH3 (depleted by a factor of 10 relative to other heavy elements) increases only by a factor of five. The pressure ranges for H2O and CH4 clouds increase by factors of 10 and 2.25, with bases reaching 650 and 0.55 bar respectively. NH3 clouds form between 4 and 6 bars only at low metallicities, while H2S clouds appear between 2 and 5 bars.

2. Elemental ratio effects: Varying C/O from 0.1 to 2.0 increases the CH4 deep mixing ratio by a factor of 3.5 and cloud thickness by a factor of 1.65. The S/N ratio controls H2S-NH3 chemistry: when S/N is below 1, NH3 is more abundant and condenses (at 3.6-8 bars), while for S/N > 1, H2S condenses instead (at 1.6-4 bars). The NH4SH cloud forms at pressures between 17 and 42 bars regardless of S/N ratio.

3. Temperature effects: A difference of 5 K in T1bar leads to 25 K difference at 100 bars and 40 K or more below 1000 bars. Cold atmospheres (T1bar = 66 K) produce thicker clouds at lower pressures, with H2O clouds at 40-510 bars, CH4 clouds at 0.32-0.75 bars, and H2S clouds at 2.6-8 bars. Warmer atmospheres (T1bar = 86 K) have thinner clouds at higher pressures, with H2O clouds at 17-180 bars and H2S clouds at 1.2-2.1 bars. CH4 does not significantly condense beyond T1bar = 71 K for a 30 solar metallicity atmosphere.

The authors identify four extreme possible atmospheric scenarios: cold heavy-element-poor atmospheres with few clouds, warm heavy-element-poor atmospheres with nearly no clouds, cold heavy-element-rich atmospheres with thick clouds, and warm heavy-element-rich atmospheres with moderate cloud formation.

Comparing with previous models, the authors find their results differ from Hueso et al. (2020), particularly for CH4 clouds which form at higher altitudes (0.1-1 bar versus 1-2 bar) and only at metallicities above 40 solar. They attribute these differences to the use of different thermodynamic datasets and the inclusion of more chemical species in FastChem.

The paper concludes that a wide range of atmospheric structures are possible for Uranus and Neptune given current observational constraints, and emphasizes the need for future ground-based observations and dedicated space missions to better characterize these planets' atmospheres.

Improvements to AI Systems:

1. Physics-constrained atmospheric retrieval models: Integrate FastChem’s thermochemical equilibrium calculations directly into Bayesian retrieval frameworks for Uranus/Neptune spectra. The improved AI can simultaneously infer metallicity, C/O, S/N ratios, and 1-bar temperature from observed spectra, while automatically enforcing consistency with cloud condensation physics (e.g., no CH4 clouds above 0.1 bar for high metallicity).

2. Uncertainty-aware cloud classification: Train a deep learning classifier on the paper’s 4 extreme atmospheric scenarios (cold/warm × poor/rich) using synthetic spectra generated from the model grid. The AI can then classify observed planetary spectra into these regimes with quantified confidence, flagging ambiguous cases that require new observations.

3. Adiabat-aware temperature profile prediction: Build a neural network that predicts full pressure-temperature profiles (0.1–1000 bar) from just T1bar and metallicity, using the Leconte et al. (2017) moist/dry adiabat formalism. This enables fast forward modeling for mission planning (e.g., probe descent trajectories) without running full radiative-convective codes.

4. Cloud deck height estimator from remote sensing: Develop a regression model that maps observable quantities (e.g., methane band depths, thermal emission slopes) to cloud base/top pressures for each species (H2O, CH4, H2S, NH3, NH4SH), using the paper’s pressure ranges as training targets. The AI can then predict cloud vertical structure from ground-based or spacecraft photometry.

5. Sensitivity analysis for observation design: Use the model’s parameter space to train an active-learning agent that recommends optimal wavelengths/filter sets for future telescopes (e.g., JWST, ELT) to discriminate between competing atmospheric structures (e.g., S/N > 1 vs < 1, or T1bar = 66 vs 86 K). The agent minimizes expected posterior entropy in retrieved parameters.

6. Cross-model consistency checker: Implement an AI that automatically compares new atmospheric models (e.g., from GCMs or other chemical codes) against FastChem results, flagging discrepancies due to thermodynamic datasets or missing species (as the paper found vs. Hueso et al. 2020). This improves reproducibility and helps identify systematic biases in exoplanet/ice giant retrievals.

7. Metallicity-dependent cloud opacity emulator: Train a surrogate model that predicts wavelength-dependent cloud opacities (from the computed cloud decks) as a function of metallicity, C/O, S/N, and T1bar. This emulator can be embedded in radiative transfer codes for rapid exploration of ice giant atmospheres, enabling real-time fitting of large spectral datasets.

What the improved AI system can do:

• Given a spectrum of Uranus/Neptune, it can simultaneously retrieve the full vertical structure (P-T profile, cloud decks, gas mixing ratios) with physical consistency, while quantifying degeneracies (e.g., between metallicity and T1bar).

• It can predict whether a planet’s atmosphere is cloud-free or has thick H2O/CH4 clouds, and at what pressures, from just a few broadband photometric points.

• It can design optimal observational campaigns (wavelengths, cadence) to break degeneracies in S/N and C/O ratios, directly informing future space missions (e.g., Uranus Orbiter and Probe).

• It can rapidly generate synthetic spectra for thousands of atmospheric scenarios, enabling statistical comparison with large survey datasets (e.g., from JWST) and identifying outlier planets with non-equilibrium chemistry.

The composition and temperature-pressure profile of the atmospheres of Uranus and Neptune are not well-determined. As observational data are limited, we often rely on chemical equilibrium computations to infer atmospheric abundances and cloud formation. The inferred atmospheric structures, however, strongly depend on several fundamental assumptions such as the elemental abundances and ratios, the condensation properties of the assumed species, or a reference temperature for the adiabatic structure. In this study we investigate the effects of different metallicities (1 to 80 solar), element ratios (C/O and S/N, from 0.1 to 2 and 0.19 to 1.6) and 1 bar temperatures (66 to 86 K) on the vertical structure of ice giant atmospheres. In particular, we use the chemical equilibrium code FastChem to derive mixing ratios and cloud structures for CH 4, NH 3, H 2 S, H 2 O and NH 4 SH. We find that the models are very sensitive to the assumed parameters, yielding drastically different possible atmospheric structures. For the cases considered here, we find that mixing ratios and cloud deck altitudes can vary by more than an order of magnitude. Additionally, thermal profiles can differ by several tens of kelvins due to composition and 1-bar temperature. We advise that future ground-based observations and a dedicated mission to Uranus and/or Neptune are required to better characterize the atmospheric structure and composition of ice giants.