Tighter constraints on the atmosphere of GJ 436 b from combined high-resolution CARMENES and CRIRES$^+$ observations

We aim to study the atmospheric properties of the warm Neptune GJ 436 b by combining a set of five transit events observed with the CARMENES spectrograph with one transit from CRIRES$^+$ so as to provide the most constrained results possible at high resolution. We removed telluric and stellar signals from the data using SysRem and potential planetary signals were investigated using the cross-correlation technique. Following standard procedures for undetected species, we performed injection recovery tests and Bayesian retrievals to place constraints on the detectability of the main near-infrared absorbers. In addition, we simulated ELT/ANDES observations by computing end-to-end in silico datasets with EXoPLORE. No molecular signals were detected in the atmosphere of GJ 436 b, which is consistent with previous studies. Combined CARMENES-CRIRES$^+$ injection-recovery and Bayesian retrieval analyses show that the atmosphere is likely covered by high-altitude clouds ($\sim$ $1$ mbar) at low and intermediate metallicities or, alternatively, is very metal-rich ($\gtrsim$ $900\times$ solar), which would suppress spectral features without invoking clouds. Simulations of ELT/ANDES observations suggest a boost by nearly an order of magnitude to the upper limit in the photon-limited regime, reaching $0.1$ mbar at $10$-$300\times$ solar metallicities. The joint analysis of all useful transit observations from CARMENES and CRIRES$^+$ provides the most stringent constraints to date on the atmospheric properties of GJ 436 b. Complementary CCF-based and retrieval approaches consistently indicate that the atmosphere is either cloudy or highly metal enriched. Any weak near-infrared absorption lines, if present, are likely to be below current detection limits. However, according to our simulations, these features may be revealed with ELT/ANDES even in single-transit observations.

💡 Research Summary

The warm Neptune GJ 436b has long been known for its featureless transmission spectrum at low to medium resolution, a result that can be explained either by high‑altitude clouds or by a highly metal‑rich, compact atmosphere. To obtain the most stringent constraints on its atmospheric composition at high spectral resolution, the authors combined five transit observations obtained with the CARMENES near‑infrared channel (R≈80 400, 0.96–1.71 µm) and one transit observed with CRIRES+ on the VLT (R≈100 000, 1.49–1.78 µm). After standard reduction, the data were processed with the SysRem algorithm (5–7 iterations) to remove telluric and stellar residuals. Cross‑correlation functions (CCFs) were then computed using high‑resolution template spectra for H₂O, CH₄, and CO, accounting for the planet’s orbital velocity and possible wind‑induced shifts. No statistically significant CCF peaks were found in any individual dataset or in the combined analysis, confirming the non‑detection of these key molecules.

To translate non‑detections into quantitative limits, the authors performed extensive injection‑recovery tests. Model spectra spanning a grid of pressures (0.1–10 mbar) and metallicities (1–3000× solar) were injected into the real data, and the recovery rate was measured after the same SysRem and CCF processing. The 3σ upper limits derived from these tests were fed into a Bayesian retrieval framework (nested sampling) that simultaneously constrained cloud top pressure, metallicity, temperature profile, and line‑broadening parameters. The posterior distributions reveal two viable atmospheric scenarios: (1) a high‑altitude cloud deck at ~1 mbar that mutes spectral features, compatible with low to intermediate metallicities (≤30× solar); or (2) a cloud‑free atmosphere that is extremely metal‑rich (≥900× solar), which compresses the scale height and suppresses line depths. Both scenarios are consistent with the CCF non‑detections and the retrieval upper limits.

Recognizing the limitations of current facilities, the authors simulated observations with the forthcoming ELT high‑resolution spectrograph ANDES using the EXoPLORE end‑to‑end pipeline. The simulations show that, in the photon‑limited regime, a single transit with ANDES could improve the detection threshold by nearly an order of magnitude, reaching pressures as low as 0.1 mbar for metallicities between 10 and 300× solar. This suggests that ELT/ANDES will be capable of distinguishing between the cloudy and ultra‑metal‑rich scenarios, and may finally reveal the weak near‑infrared absorption lines that are currently below detection limits.

In summary, the joint CARMENES‑CRIRES+ analysis provides the most stringent high‑resolution constraints on GJ 436b to date: no H₂O, CH₄, or CO is detected, and the atmosphere is either cloud‑covered at ~1 mbar or possesses a metallicity exceeding 900× solar. The study demonstrates the power of combining multiple high‑resolution transit datasets, injection‑recovery testing, and Bayesian retrievals, and it highlights the transformative potential of next‑generation ELT spectroscopy for characterizing cloudy sub‑Neptune exoplanets.