pith. machine review for the scientific record. sign in

arxiv: 2604.15292 · v1 · submitted 2026-04-16 · 🌌 astro-ph.EP

Recognition: unknown

Cloudy with a chance of metals: Indications of CO₂ in the atmosphere of GJ 1214 b from high-resolution K-band spectroscopy

L. Nortmann , D. Cont , F. Lesjak , A. D. Rains , A. Lavail , L. Boldt-Christmas , E. Nagel , A. Reiners
show 8 more authors
N. Piskunov F. Yan A. Hatzes O. Kochukhov D. Shulyak U. Seemann M. Rengel A. Hahlin
Authors on Pith no claims yet

Pith reviewed 2026-05-10 09:41 UTC · model grok-4.3

classification 🌌 astro-ph.EP
keywords exoplanet atmospherestransmission spectroscopyGJ 1214 bcarbon dioxidehigh-resolution spectroscopysub-Neptuneatmospheric retrieval
0
0 comments X

The pith

High-resolution K-band spectroscopy indicates CO2 absorption in the atmosphere of GJ 1214 b.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents observations of eight transits of the sub-Neptune GJ 1214 b with the CRIRES+ spectrograph in the K band. Telluric and stellar signals were removed with SYSREM, after which cross-correlation searches yielded non-detections for H2O, CO, CH4, H2S, and NH3, from which upper limits were placed on their abundances. A cross-correlation signal for CO2 reached S/N of 3.6, with a Welch t-test showing the in-trail and out-of-trail distributions differ at 3.4 sigma. Bayesian retrievals returned a metallicity of [M/H] = 0.48, cloud deck pressure log10(Pc) = -3.04, and isothermal temperature 398 K, values consistent with JWST data within 1.5 sigma and implying CO2 is a significant atmospheric component.

Core claim

The central claim is that a cross-correlation function peak for CO2 appears at S/N ~ 3.6 in the processed transmission spectra, with no clear evidence of origin in correlated noise and a t-test confirming 3.4 sigma separation of in-trail versus out-of-trail values. Retrieval under free chemistry yields volume mixing ratios that correspond to super-solar metallicity, a high-altitude opacity deck, and a temperature intermediate between the day- and night-side profiles from JWST, all compatible with the observed K-band data and prior mid-infrared constraints.

What carries the argument

Cross-correlation of template spectra for individual molecules against the SYSREM-detrended high-resolution transmission spectra to extract molecular signatures.

If this is right

  • CO2 must be included at appreciable abundance when modeling the transmission spectrum of GJ 1214 b at both high and low spectral resolution.
  • Upper limits on the other searched molecules constrain the possible range of carbon and nitrogen chemistry in the atmosphere.
  • The retrieved parameters remain compatible with JWST NIRSpec data within the models' stated uncertainties.
  • Further high-resolution observations or mid-infrared data are required to raise the detection significance and tighten compositional constraints.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the CO2 signal holds, ground-based high-resolution spectroscopy can extract compositional information from sub-Neptunes even when space-based low-resolution spectra appear muted.
  • Atmospheric formation models for GJ 1214 b would need to accommodate significant CO2 enrichment, possibly linked to the planet's overall metallicity.
  • Targeted follow-up at wavelengths where CO2 bands are stronger could provide an independent test of the K-band result.

Load-bearing premise

The detected cross-correlation peak for CO2 is produced by planetary absorption rather than residual correlated noise or over-subtraction artifacts from the data processing steps.

What would settle it

Recovery of a higher-S/N CO2 signal in an independent set of transits observed with the same or another high-resolution instrument, or consistent non-recovery in additional data of comparable quality.

Figures

Figures reproduced from arXiv: 2604.15292 by A. D. Rains, A. Hahlin, A. Hatzes, A. Lavail, A. Reiners, D. Cont, D. Shulyak, E. Nagel, F. Lesjak, F. Yan, L. Boldt-Christmas, L. Nortmann, M. Rengel, N. Piskunov, O. Kochukhov, U. Seemann.

Figure 1
Figure 1. Figure 1: Conditions during the eight observations of GJ 1214 b. Top to bottom: Mean S/N per pixel for each spectrum, the progression of air￾mass, the relative humidity, and the absolute radial velocity shift (∆RV) between the planet and Earth rest frame are shown. The latter consist of the sum of the barycentric velocity, which changes during the observa￾tion, and the static system velocity. The transit window is i… view at source ↗
Figure 2
Figure 2. Figure 2: Stages of the data after progressive analysis steps, shown on the example of the third wavelength segment (2002.0-2014.6 nm) of night 5. Top to bottom: Data are shown as 1. the unnormalised time series; 2. after normalisation; 3. after masking of the deepest telluric lines (≤ 20% of the continuum level); 4. after nine SYSREM iterations. The lowest two panels show the same stage of analysis as the fourth; 5… view at source ↗
Figure 3
Figure 3. Figure 3: Effect of cloud deck height and atmospheric metallicity on planetary transmission spectra. Left panel: Effect of increasing height of the cloud deck, by decreasing the grey cloud deck pressure, Pc , of the models is illustrated on a model containing only CH4 lines. Right panel: Effect of decreasing atmospheric scale height, caused by increasing atmospheric metallicity [M/H], and thus the mean molecular wei… view at source ↗
Figure 4
Figure 4. Figure 4: Results of the cross-correlation analysis. Left column: Synthetic model templates used for the individual molecules. Middle column: Cross￾correlation functions for every spectrum of all eight nights, sorted in phase and shifted to planet rest frame. The expected position of a signal should be vertical at v = 0 km s−1 between the two horizontal dashed lines which indicate the first and last point of contact… view at source ↗
Figure 5
Figure 5. Figure 5: Results of the S/N grid analysis that show the recovered detection significance of injected synthetic models with varying metallicities [M/H] and opacity deck pressures, Pc . The injected models contained CIA continuum absorption and lines of one molecule (H2O, CH4, CO, CO2, H2S, and NH3, respectively). The white lines represent the contours of S/N=3 (dotted) and S/N=4 (solid) [PITH_FULL_IMAGE:figures/ful… view at source ↗
Figure 6
Figure 6. Figure 6: Decomposition of the CO2 signal into its individual contribu￾tions. Top panel: CO2 template spectrum (dark blue) and the 18 wave￾length segments covered by the K2148 setting (shaded regions). Bottom panel: Decomposition of the S/N=3.6 peak signal of the CO2 CCF into its contributions from individual nights and wavelength segments given in terms of the per cent of the cumulative signal. Wavelength segments … view at source ↗
Figure 7
Figure 7. Figure 7: Full Kp - v map of GJ 1214 b. Left panel: calculated with a CO2 model at [M/H]=1.5 and no clouds normalised and shown with extending the normalisation range to the negative Kp range, and excluding nights 1 and 2. Right panel: Same as left but zoomed in to the peak CO2 signal. The dotted white lines indicate literature Kp (horizontal) and v = 0 (vertical). our exploration and consequently the normalisation … view at source ↗
Figure 8
Figure 8. Figure 8: Isothermal temperature-pressure profile retrieved in the free chemistry retrieval. Its 1σ uncertainty intervals are indicated in blue shaded regions. The profile is compared to the profiles and their 1σ un￾certainties retrieved by Kempton et al. (2023) from JWST MIRI data of day- and night-side emission. The latter were extracted from their [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Upper limits for six molecules from the free chemistry retrieval. We show the posterior distributions and mark in dark blue the 2σ upper limits (calculated as the 97.5th percentile). The corresponding value is given above the histograms. In orange we mark the median and 1σ uncertainty intervals, also listed in [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of the multi-molecule, free-chemistry retrieval models to low-resolution data. Top panel: Predicted relative transit depth change from the free chemistry models retrieved from CRIRES+ data (green). The model predictions are plotted in high-resolution in the wave￾length region overlapping with the CRIRES+ remeasurements indicated by the grey shaded areas and in low-resolution for the rest of the… view at source ↗
Figure 11
Figure 11. Figure 11: Simulated S/N grids showing the expected constraints that can be obtained by obtaining more data. The dotted white lines represent the contours of the simulated data, with S/N=3 (dotted) and S/N=4 (solid) respectively. The dark blue lines represent the contours of the injection recovery tests of our real eight transit data set shown in [PITH_FULL_IMAGE:figures/full_fig_p015_11.png] view at source ↗
read the original abstract

Sub-Neptune exoplanets frequently exhibit muted transmission spectra, with GJ 1214 b being the most prominent example. Following years of intense observing campaigns yielding featureless planetary spectra, recent observations with JWST revealed the first possible atmospheric signatures. We present high-resolution transmission spectroscopy of GJ 1214 b based on eight transits obtained with the CRIRES$^+$ spectrograph in the K band. We used SYSREM to remove telluric and stellar signals and searched for signatures of H2O, CO, CH4, H2S, NH3, and CO2 using the cross-correlation technique. We obtained non-detections for the first five molecules and used injection recovery tests to derive upper limits on the atmosphere. For CO$_2$ we measure a CCF signal at S/N ~ 3.6, with a detailed investigation showing no obvious indication that it is caused by correlated noise. A Welch t-test confirmed the in-trail and out-of-trail distributions to be different at $3.4 \sigma$ confidence. A Bayesian retrieval framework with free chemistry, resulted in volume mixing ratios corresponding to a metallicity of $[\mathrm{M/H}]=0.48^{+0.89}_{-1.70}$, an opacity deck pressure of $\log_{10}(P_\mathrm{c}) = -3.04^{+2.52}_{-1.53}$ and a planet temperature of $T_\mathrm{iso}=398^{+283}_{-197}$ K, consistent with a value intermediate between the day- and night-side T-p's derived from JWST data. While these values correspond to relatively large signal amplitudes predicted for CO2 features in the mid-infrared, they are compatible with JWST NIRSpec observations within the models' $1.5\sigma$ uncertainties. Further modelling and additional data are required to confirm the atmospheric signatures and obtain a comprehensive interpretation of low- and high-resolution data. Overall, our results support previous findings that CO2 is likely to be a significant component of the atmosphere of GJ 1214 b.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The manuscript presents high-resolution K-band transmission spectroscopy of the sub-Neptune GJ 1214 b using eight transits observed with CRIRES+. After SYSREM detrending to remove telluric and stellar lines, cross-correlation functions (CCF) are computed for H2O, CO, CH4, H2S, NH3, and CO2. Non-detections are reported for the first five species with derived upper limits via injection tests; for CO2 a CCF peak at S/N ~3.6 is found, supported by a Welch t-test showing in-trail vs. out-of-trail distributions differ at 3.4σ. A free-chemistry Bayesian retrieval yields [M/H] = 0.48^{+0.89}_{-1.70}, log10(Pc) = -3.04^{+2.52}_{-1.53}, and T_iso = 398^{+283}_{-197} K, stated to be consistent with JWST-derived T-p profiles within 1.5σ. The authors conclude that CO2 is likely a significant atmospheric constituent.

Significance. If the CO2 detection holds, the result would supply the first ground-based high-resolution indication of CO2 in a sub-Neptune atmosphere and demonstrate complementarity between CRIRES+ CCF signals and JWST low-resolution spectra. It would also constrain cloud-deck pressure and metallicity for GJ 1214 b, reinforcing the emerging picture that CO2 may dominate the observable features in this class of planets. The work is technically timely given the recent JWST data, but the marginal statistical significance limits its immediate weight in the literature.

major comments (2)
  1. [CO2 CCF detection and statistical tests] In the CO2 cross-correlation results: the reported S/N ~3.6 peak and 3.4σ Welch t-test are marginal. The central claim that this constitutes an atmospheric detection rests on the assumption that SYSREM has neither attenuated a real signal nor introduced correlated residuals; without quantitative robustness checks (e.g., CCF behavior across a range of SYSREM iterations or false-alarm rates from phase-shuffled or null datasets), the distinction between planetary absorption and artifact remains uncertain and load-bearing for the paper's conclusion.
  2. [Bayesian retrieval framework] In the atmospheric retrieval section: the Bayesian fit is performed on the identical dataset that produced the CCF peak, so the retrieved parameters ([M/H], Pc, T_iso) cannot be regarded as an independent confirmation. The posteriors are extremely broad (metallicity spanning >2 dex, Pc and T_iso uncertainties of order 2–3 dex/K), weakening any claim that the retrieval supports the CO2 identification or provides meaningful consistency with JWST.
minor comments (3)
  1. The abstract states that the retrieval 'resulted in' the quoted parameters; rephrase for grammatical clarity.
  2. [Cross-correlation methodology] Clarify the precise definition and normalization used for the CCF S/N value (peak height relative to what noise estimate?) and whether it accounts for the number of trials across molecules.
  3. The injection-recovery tests for upper limits on non-detected species are mentioned but not quantified in the provided text; a table or figure summarizing the recovered S/N versus injected VMR would strengthen the non-detection claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments, which help clarify the statistical limitations of our analysis. We address each major point below and propose revisions to improve the robustness and clarity of the manuscript.

read point-by-point responses

  1. Referee: [CO2 CCF detection and statistical tests] In the CO2 cross-correlation results: the reported S/N ~3.6 peak and 3.4σ Welch t-test are marginal. The central claim that this constitutes an atmospheric detection rests on the assumption that SYSREM has neither attenuated a real signal nor introduced correlated residuals; without quantitative robustness checks (e.g., CCF behavior across a range of SYSREM iterations or false-alarm rates from phase-shuffled or null datasets), the distinction between planetary absorption and artifact remains uncertain and load-bearing for the paper's conclusion.

    Authors: We agree that the S/N of 3.6 is marginal and that the detection claim depends on SYSREM not having introduced or removed correlated features. The manuscript already includes a detailed investigation ruling out obvious correlated noise artifacts, but we did not quantify behavior across SYSREM iterations or compute false-alarm rates via phase shuffling. We will add these quantitative checks (including CCF maps for a range of SYSREM iterations and false-alarm probabilities from phase-shuffled and null datasets) to a revised version, which will allow readers to better assess the signal reliability. revision: yes

  2. Referee: [Bayesian retrieval framework] In the atmospheric retrieval section: the Bayesian fit is performed on the identical dataset that produced the CCF peak, so the retrieved parameters ([M/H], Pc, T_iso) cannot be regarded as an independent confirmation. The posteriors are extremely broad (metallicity spanning >2 dex, Pc and T_iso uncertainties of order 2–3 dex/K), weakening any claim that the retrieval supports the CO2 identification or provides meaningful consistency with JWST.

    Authors: We acknowledge that the retrieval uses the same dataset as the CCF analysis and therefore does not provide independent confirmation of the detection. Its purpose is to derive posterior constraints on atmospheric parameters given the observed signal and to test consistency with JWST-derived T-p profiles. The broad posteriors are expected given the marginal signal strength and known degeneracies between metallicity, cloud pressure, and temperature; we already report the 1.5σ consistency with JWST. We will revise the text to explicitly state that the retrieval offers supporting parameter estimates rather than independent verification, while retaining the reported consistency metric. revision: yes

Circularity Check

0 steps flagged

No significant circularity; analysis chain is self-contained against external benchmarks

full rationale

The paper applies standard SYSREM detrending and cross-correlation to high-resolution spectra to search for molecular signatures, reports a marginal CO2 CCF detection with statistical tests, then performs a Bayesian retrieval on the same spectra to derive atmospheric parameters. These parameters are compared to independent JWST results rather than being presented as predictions derived from the fit itself. No equations or steps reduce by construction to the inputs, no self-citation chains support load-bearing claims, and no ansatzes or uniqueness theorems are invoked from prior author work. The derivation remains independent of its own fitted outputs.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The claim depends on fitted retrieval parameters and on the assumption that the marginal CCF peak originates from the planet rather than systematics.

free parameters (3)
  • metallicity [M/H]
    Fitted parameter in the free-chemistry Bayesian retrieval.
  • opacity deck pressure log10(Pc)
    Fitted parameter in the free-chemistry Bayesian retrieval.
  • isothermal temperature T_iso
    Fitted parameter in the free-chemistry Bayesian retrieval.
axioms (2)
  • domain assumption SYSREM removes telluric and stellar signals without attenuating the planetary CO2 signature.
    Invoked in the data reduction step described in the abstract.
  • domain assumption The 3.6 S/N CCF peak is not produced by correlated noise.
    Central assumption required to interpret the signal as planetary; only partially addressed by the Welch test.

pith-pipeline@v0.9.0 · 5767 in / 1631 out tokens · 51153 ms · 2026-05-10T09:41:40.711342+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

11 extracted references · 2 canonical work pages

  1. [1]

    Adams, D., Gao, P., de Pater, I., & Morley, C. V . 2019, ApJ, 874, 61 Allart, R., Lemée-Joliecoeur, P.-B., Jaziri, A. Y ., et al. 2023, A&A, 677, A164 Alonso-Floriano, F. J., Sánchez-López, A., Snellen, I. A. G., et al. 2019, A&A, 621, A74 Banerjee, A., Barstow, J. K., Gressier, A., et al. 2024, ApJ, 975, L11 Bean, J. L., Désert, J.-M., Kabath, P., et al....

    work page arXiv 2019

  2. [2]

    Herman et al

    or, the more model agnostic approach, to identify the iteration after which the change in the standard deviation of the residu- als from one iteration to the next reaches a plateau (e.g. Herman et al. 2020, 2022; Deibert et al. 2021; Ridden-Harper et al. 2023; Lesjak et al. 2025). Keeping the number ofSYSREMiterations low is considered desirable as progre...

    work page arXiv 2020

  3. [3]

    Wenger et al

    RV semi-amplitudeK ⋆ (m s−1) 14.36±0.53 (2) References.1. Wenger et al. (2000),

    2000

  4. [4]

    Cloutier et al. (2021),

    2021

  5. [5]

    Kempton et al. (2023),

    2023

  6. [6]

    Charbonneau et al. (2009),

    2009

  7. [7]

    (2020) (the latter work corrects the sign of the value given in 5.) differences in our respective data analysis

    Kasper et al. (2020) (the latter work corrects the sign of the value given in 5.) differences in our respective data analysis. To filter the model we needed to bring it into the same format as the data, that is, for each of the 18 wavelength segments a 2D matrix with the two dimensions being time steps (number of exposures in the re- spective night) and w...

    2020

  8. [8]

    M′ =M−Cinstead, as was done by Gibson et al

    Since we subtracted eachSYSREMiteration during the iteration process in which we obtained the cumula- tive correction matrix, but then divided instead of subtracted this correction matrix from the data, we also divide our model by our correction term to obtain the filtered model matrix M′ =(M+1)/(C+1)−1.(A.3) We note that applying the correction by subtra...

    2022

  9. [9]

    1 and Sect

    We omitted simulations of nights 1 and 2 due to the unfavourable total velocity shift between the planet lines and telluric lines in these nights (see Fig. 1 and Sect. 4.3.1). Our transmission spec- Article number, page 20 of 24 L. Nortmann et al.: Indications of CO 2 in the atmosphere of GJ 1214 b from high-resolution K-band spectroscopy Fig. A.7.K p-vma...

    2025

  10. [10]

    (2019) and using the stellar parameters from Cloutier et al

    model spectra first described by Nordlander et al. (2019) and using the stellar parameters from Cloutier et al. (2021). Our telluric spectra were generated from the K band per-molecule (H2O, CH4, N2O, CO2, CO) template transmission spectra computed by Köhler et al. (2025) for thevipersoft- ware package, and we allowed H 2O absorption and slit losses to sm...

    2019

  11. [11]

    aa58512-25corr_appliedArXiv Fig

    Article number, page 21 of 24 A&A proofs:manuscript no. aa58512-25corr_appliedArXiv Fig. A.8.Results of the Bayesian retrieval run on our GJ 1214 b high-resolution data. The free parameters of the retrieval were the planet’s temperature,T iso, used to calculate an isothermalT-pprofile, the opacity deck top pressure, log 10(Pc), the C/O ratio and metallici...

    2024