🔭 astrophysics

A parameterised approach to disequilibrium retrievals in the JWST era: Application to NIRCam observations of HD 189733b

This paper introduces a parameterised retrieval framework that accounts for transport-induced disequilibrium chemistry, demonstrating its ability to avoid biases in atmospheric parameters and successfully applying it to JWST/NIRCam observations of HD 189733b to provide the first tentative constraints on quenching and photochemical activity in a hot Jupiter atmosphere.

Original authors: Jake Taylor, Shang-Min Tsai, Vivien Parmentier, Chloe Fisher, Michael Line

Published 2026-07-08

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jake Taylor, Shang-Min Tsai, Vivien Parmentier, Chloe Fisher, Michael Line

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Reading the Sky's Recipe Book

Imagine you are trying to figure out what a cake is made of just by looking at a picture of it. In astronomy, scientists do this with planets. They look at the light passing through a planet's atmosphere (like a "transit") to see which chemical "ingredients" are there. This process is called atmospheric retrieval.

For a long time, scientists assumed these planetary atmospheres were like a perfectly stirred pot of soup: everything was mixed evenly and settled into a stable, predictable state called thermochemical equilibrium. They used pre-made "recipe cards" based on this assumption to guess the planet's temperature, metal content, and chemical ratios.

The Problem: Real atmospheres aren't calm pots of soup. They are like a stormy ocean with strong currents. In hot Jupiters (giant, hot planets), strong winds mix the air from deep, hot layers up to the cold, high layers very quickly. This "vertical mixing" stops the chemicals from settling into their natural, stable state. If you use the old "calm soup" recipe cards on a "stormy ocean" planet, you get the wrong recipe. You might think the planet is made of different ingredients than it actually is.

What This Paper Did: The "Quench" Switch

The authors, led by Jake Taylor, wanted to fix this. They built a new tool to analyze data from the James Webb Space Telescope (JWST), specifically looking at a planet called HD 189733 b.

Instead of assuming the atmosphere is calm, they introduced a "switch" called quenching.

The Analogy: Imagine a deep-sea diver (the atmosphere) who is swimming up very fast. As they rise, the water pressure drops. If they swim up too fast, the air bubbles in their lungs don't have time to adjust; they get "frozen" or "quenched" at the pressure where they started rising.
In the Paper: The authors added a new variable to their math: the "quench pressure." This is the specific depth in the atmosphere where the winds are so fast that they freeze the chemical mix in place, carrying it up to the surface before it can change.

They tested two ways to do this:

Elemental Quenching: Grouping chemicals by their family (e.g., all carbon-based molecules get frozen at the same depth).
Molecular Quenching: Freezing each specific molecule (e.g., Methane gets frozen at one depth, Ammonia at another).

The Test Drive: Synthetic Data

Before looking at real data, they created "fake" JWST observations (synthetic data) of HD 189733 b. They made these fake atmospheres with different wind speeds (mixing strengths) and then tried to decode them using three different methods:

The Old Way: Assuming everything is in equilibrium (no mixing).
The New Way (Elemental): Allowing for mixing.
The New Way (Molecular): Allowing for mixing, molecule by molecule.

The Result:

When they used the Old Way on a planet with strong winds, the results were wrong. They guessed the wrong amount of carbon and metal, like trying to guess a cake's ingredients while ignoring that the frosting was smeared everywhere.
When they used the New Way, they successfully recovered the correct ingredients, even when the winds were strong. They proved that by just adding a couple of extra "knobs" (the quench pressures) to their model, they could see the truth behind the storm.

The Real Deal: Analyzing HD 189733 b

Finally, they applied their new tool to real data from JWST.

1. The Methane Mystery:
Previous studies noticed that this planet had very little Methane (CH4), which was strange. The authors found that their new model explained this perfectly: the winds were so strong that they froze the Methane deep in the atmosphere where it is scarce, and carried that "low methane" mix up to the surface. The planet wasn't missing methane; it was just being carried up from a deep, dry layer.

2. The H2S Discovery (The New Ingredient):
The paper introduces a clever new trick to study Hydrogen Sulfide (H2S).

The Analogy: Imagine a light switch in a room. Below the switch, the room is dark (deep atmosphere). Above the switch, the light turns on (photochemistry), and things start to change.
The Discovery: They created a model that assumes H2S is mixed evenly deep down, but once it reaches a certain height (the "break pressure"), sunlight starts breaking it apart.
The Result: They found tentative evidence that this "light switch" exists in the atmosphere of HD 189733 b. This is the first time scientists have been able to estimate where the photochemical (sunlight-driven) region begins in a hot Jupiter's atmosphere.

The Takeaway

This paper shows that to understand exoplanets correctly, we can't just assume their atmospheres are calm and stable. We have to account for the "storms" (vertical mixing) that freeze chemicals in place.

By adding a simple "quench pressure" setting to our models, we can:

Get the correct chemical recipes (C/O ratios and metallicity).
Explain why certain gases (like Methane) seem missing.
Map out where sunlight starts to change the atmosphere (the H2S photochemical region).

The authors conclude that future studies of exoplanet atmospheres should include these "quench" settings to avoid getting the recipe wrong.

Technical Summary: A Parameterised Approach to Disequilibrium Retrievals in the JWST Era

Problem Statement
Atmospheric retrievals are essential for inferring the physical and chemical properties of exoplanets from observed spectra. However, a common simplifying assumption in these analyses is that the atmosphere exists in thermochemical equilibrium. While this assumption holds for highly irradiated hot Jupiters, it often fails for cooler objects or those with strong vertical mixing, where transport-induced quenching drives chemical abundances out of equilibrium. Assuming equilibrium in the presence of disequilibrium can lead to significant biases in retrieved bulk atmospheric parameters, specifically the carbon-to-oxygen ratio (C/O) and metallicity ([M/H]). Existing retrieval frameworks face a trade-off: "free-chemistry" models are flexible but prone to overfitting; equilibrium models are efficient but physically inaccurate for mixing scenarios; and full radiative-convective-disequilibrium models are physically consistent but computationally expensive and reliant on pre-computed grids.

Methodology
This study presents a parameterised retrieval framework that bridges the gap between efficiency and physical realism by explicitly modeling transport-induced quenching. The authors utilize the NemesisPy radiative transfer and retrieval tool, coupled with FastChem 3.0 for equilibrium chemistry and VULCAN for generating synthetic disequilibrium abundance profiles.

Synthetic Data Generation: The authors generated synthetic James Webb Space Telescope (JWST) transmission spectra for the hot Jupiter HD 189733 b. These spectra were produced using VULCAN to simulate chemical abundance profiles under varying vertical mixing strengths, parameterized by the eddy diffusion coefficient ( $K_{zz} = 10^6, 10^8, 10^{10}, 10^{12} \text{ cm}^2\text{s}^{-1}$ ).
Disequilibrium Parameterisation: To incorporate disequilibrium effects without the computational cost of full kinetic grids, the authors introduced free parameters for quench pressures. Two approaches were tested:
- Elemental Quenching: Grouping species by elemental constituents (e.g., quenching all carbon-bearing species at a single pressure $P_{q,C}$ and nitrogen-bearing species at $P_{q,N}$ ).
- Molecular Quenching: Treating individual molecules (e.g., $\text{CH}_4$ , $\text{NH}_3$ ) as having independent quench pressures.
  The abundance profile for a species $X$ is defined as the equilibrium value below the quench pressure ( $P > P_{\text{quench}}$ ) and constant at the equilibrium value calculated at $P_{\text{quench}}$ above it ( $P \le P_{\text{quench}}$ ).
Retrieval Simulations: The framework was tested against synthetic observations from JWST instruments NIRISS/SOSS and NIRSpec/G395H. Retrievals were performed assuming: (i) thermochemical equilibrium, (ii) elemental quenching, and (iii) molecular quenching.
Application to Real Data: The framework was applied to published JWST/NIRCam transmission observations of HD 189733 b (Fu et al. 2024). A novel $\text{H}_2\text{S}$ parameterisation was developed to model the photochemically active region, defining a deep abundance and a break pressure ( $P_{\text{break}}$ ) above which abundance declines via a power law.

Key Results

Bias in Equilibrium Assumptions: Synthetic retrieval tests demonstrated that assuming thermochemical equilibrium when vertical mixing is present leads to significant biases. Specifically, equilibrium retrievals often underestimated C/O and [M/H] values, failing to recover the true input parameters as mixing strength increased.
Recovery of Bulk Properties: The disequilibrium retrieval frameworks (both elemental and molecular) successfully recovered the correct bulk atmospheric properties (C/O and [M/H]) in most synthetic cases, even when strong vertical mixing was present. The inclusion of just two additional free parameters (quench pressures) was sufficient to correct the biases inherent in equilibrium models.
Quench Pressure Constraints: The framework successfully retrieved the approximate quenching pressures for carbon and nitrogen species. Results indicated that nitrogen chemistry quenches at deeper atmospheric levels than carbon chemistry, consistent with theoretical expectations regarding reaction timescales.
Application to HD 189733 b: When applied to real JWST data, the disequilibrium model provided a better fit (higher Bayesian evidence) than the chemical equilibrium model. The best-fitting model suggested strong vertical mixing with a carbon quench pressure of $\log P_{q,C} \approx 1.7$ bar.
$\text{H}_2\text{S}$ and Photochemistry: The study provided tentative evidence for the photochemically active region of the atmosphere. By parameterizing the $\text{H}_2\text{S}$ vertical profile, the authors constrained the break pressure ( $P_{\text{break}}$ ) to $\log P_{\text{break}} \approx -3.25$ bar. This represents the first tentative constraint on the vertical gradient of an atmospheric gas in a hot Jupiter atmosphere, offering an indirect method to infer the location where atomic sulfur becomes dominant.

Significance and Claims
The paper claims to offer a computationally efficient and physically motivated alternative to both free-chemistry and full-grid retrieval methods. By parameterizing quench pressures, the authors demonstrate that it is possible to:

Correctly infer bulk atmospheric properties (C/O and metallicity) even in the presence of significant vertical mixing, avoiding the biases of equilibrium assumptions.
Probe the deep thermal structure and mixing strengths of exoplanet atmospheres through the retrieval of quench pressures.
Constrain the photochemically active regions of exoplanet atmospheres using JWST data, specifically through the novel parameterisation of $\text{H}_2\text{S}$ .

The authors conclude that future retrieval studies of exoplanet atmospheres should consider including quench pressure parameters for key molecular species to achieve a more physically accurate characterization of atmospheric composition and to account for unknown vertical mixing processes. They note that while their disequilibrium and equilibrium-with- $\text{H}_2\text{S}$ models produce similar fits to current data, the degeneracy between them could be broken by future observations with NIRISS/SOSS.