Possible stratospheric emission in the warm Neptune GJ 436 b from high-resolution spectroscopy

We present high spectral resolution $L$ band (2.91–3.85 $μ$m) observations of the warm Neptune GJ 436 b from Keck II/KPIC. KPIC’s single-mode fiber feed reduces the $L$ band sky background by a factor of 100, significantly improving sensitivity compared to a seeing-limited spectrometer and enabling a tentative ($\rm SNR = 3-4$) cross-correlation detection of GJ 436 b with a thermally inverted atmospheric model. In contrast with recent results from $JWST$ and high-resolution transmission spectroscopy, our retrieval analysis prefers the presence of H$_2$O, and possibly CH$4$, molecular features in emission. The broad-band continuum flux associated with the maximum-likelihood model is substantially higher than expected based on both the $\sim670\rm\ K$ equilibrium temperature of GJ 436 b and previous results from low-resolution spectroscopy. We demonstrate that the loss of continuum information during the processing of high-resolution spectra makes our analysis effectively insensitive to the absolute continuum level of the planet, and that scaling the maximum-likelihood model to match the broad-band flux measured from low-resolution observations of GJ 436 b results in a detection of similar strength in cross-correlation. These results could be explained by a thermal inversion arising above a haze layer in the upper atmosphere of \gjb. Further observations, ideally post-eclipse in order to break the $K_p - Δv{sys}$ degeneracy, are needed to clarify this possible detection. This work demonstrates the potential of $L$ band high-resolution spectroscopy for characterizing significantly smaller and cooler exoplanets compared with hot Jupiters.

💡 Research Summary

In this study the authors present the first L‑band (2.91–3.85 µm) high‑resolution emission spectroscopy of the warm Neptune GJ 436b using the Keck II/KPIC instrument. KPIC’s single‑mode fiber feed suppresses the thermal sky background by roughly a factor of 100 relative to a conventional seeing‑limited spectrograph, enabling a deep set of 60‑second exposures that achieve per‑pixel signal‑to‑noise ratios of 30–100 across the band. Observations were taken near secondary eclipse (orbital phase 0.42–0.48) where the planet’s line‑of‑sight velocity varies from ~89 km s⁻¹ to ~63 km s⁻¹, providing enough Doppler shift to separate planetary lines from stellar and telluric features.

Data reduction employed an ABBA nodding pattern for background subtraction, manual wavelength calibration using NIRSPEC EFS simulations followed by fine‑tuning with stellar‑telluric models, and a 5th‑order Gaussian‑Hermite line‑spread‑function (LSF) fit that yields a resolving power of R≈29 000 at the blue end and R≈21 000 at the red end (velocity resolution 10–14 km s⁻¹). After extracting the spectra, the authors removed time‑varying telluric and instrumental systematics with principal‑component analysis (PCA), discarding 2, 4, 6, or 8 components in separate runs and re‑injecting the removed components into the forward model to preserve any planetary signal.

Atmospheric retrieval was performed with a pipeline that couples petitRADTRANS radiative transfer to the Brogi & Line (2019) log‑likelihood mapping, sampled with PyMultiNest. The pressure‑temperature profile follows the Guillot (2010) parameterisation with four free parameters (log κ, log γ, T_int, T_eq). A gray cloud deck is included, with cloud pressure, opacity, and sedimentation efficiency (f_SED) as free parameters. The orbital velocity model incorporates the full Keplerian solution (including eccentricity, argument of periastron, and barycentric corrections) and treats the planet’s radial‑velocity semi‑amplitude Kp and a systemic offset Δv_sys as free parameters. Volume‑mixing ratios for H₂O, CH₄, NH₃, H₂S, HCN, SO₂, and H₂ are retrieved using up‑to‑date line lists (POKAZATEL for water, ExoMol for methane, etc.) and CIA opacities.

Two families of solutions emerge. A non‑inverted (isothermal or decreasing temperature with altitude) model reproduces the low‑resolution JWST and Spitzer data but yields a lower log‑likelihood. A thermally inverted model, featuring a temperature rise above ~10 mbar, provides a statistically significant improvement (≈3–4 σ) in the cross‑correlation function (CCF). This inverted model predicts strong emission features from H₂O and possibly CH₄, and the CCF peaks at Kp≈117 km s⁻¹ and Δv_sys≈0 km s⁻¹ with a signal‑to‑noise ratio of 3–4. The associated broadband continuum flux is substantially higher than the ∼670 K equilibrium temperature would suggest and exceeds the flux measured by previous low‑resolution observations by a factor of two to three.

The authors emphasize that high‑resolution processing removes the absolute continuum level; the spectra are normalized and the continuum is effectively lost. Consequently, the retrieved model can be arbitrarily scaled in flux without affecting the line contrast that drives the CCF detection. When the model is rescaled to match the broadband flux from low‑resolution data, the CCF detection strength remains unchanged, confirming that the detection is driven purely by line contrast rather than absolute flux.

Key limitations are identified. The Kp–Δv_sys degeneracy (the “Kp–Δv” degeneracy) prevents a unique determination of the planet’s orbital velocity and systemic offset because the observations cover only a portion of the orbit and lack post‑eclipse data that would break the symmetry. The PCA removal may also attenuate the planetary signal, potentially lowering the measured SNR. Moreover, the lack of absolute continuum information means that the claim of a “strong emission” spectrum is contingent on an assumed scaling, not a direct measurement.

Despite these caveats, the work demonstrates that L‑band high‑resolution spectroscopy can probe atmospheric layers above clouds/hazes in relatively cool (∼670 K) Neptune‑size planets—regimes previously accessible only to hot Jupiters. The detection of a possible stratospheric temperature inversion, inferred from emission lines of water and methane, offers a new avenue to study disequilibrium chemistry and vertical mixing in warm Neptunes. The authors recommend follow‑up observations that include post‑eclipse phases to resolve the Kp–Δv_sys degeneracy, longer time baselines to improve SNR, and complementary wavelength coverage (e.g., M‑band) to test the robustness of the inversion signal.

In summary, this paper provides a tentative but compelling high‑resolution detection of thermal emission from GJ 436b, suggests the presence of a high‑altitude temperature inversion possibly above a haze layer, and highlights the promise of L‑band HRCCS for characterizing smaller, cooler exoplanets beyond the hot‑Jupiter regime. Future observations will be essential to confirm the inversion, refine molecular abundances, and fully exploit the capabilities demonstrated here.