A Detailed Investigation of HD 209458 b HST & JWST Transmission Spectra with SANSAR

TL;DR

HD 209458 b's atmospheric constraints from HST and JWST vary with retrieval setup; using SANSAR, the paper demonstrates that free retrievals yield $χ^2_{\mathrm{red}}$ values around 1.21 and imply highly sub-solar metallicity with highly sub-solar C/O, while equilibrium chemistry and grid retrievals tend toward solar-like values on the same dataset. The study shows that NIRCam-only analyses can overestimate abundances, whereas incorporating UV/optical data (STIS) yields robust solar-like metallicities and C/O, and CO2 detections require optical baselines for quantitative constraints. Benchmarking on WASP-96 b demonstrates that correlated-k at modest spectral resolution reproduces line-by-line results within tens of ppm and agrees with other codes, validating SANSAR's multi-method approach. Overall, the work provides guidance for JWST-era atmospheric retrievals and highlights the importance of data combination and model choice in robustly constraining exoplanetary atmospheres.

Abstract

HD 209458 b is the first exoplanet on which an atmosphere was detected. Since then, its atmosphere has been investigated using multiple telescopes and instruments. However, many of its atmospheric constraints remain debatable. While HST observations suggested a highly sub-solar metallicity, recent JWST NIRCam observations by Xue et al. 2024 constrained a super-solar metallicity with highly sub-solar C/O. In this work, we show a detailed investigation of HD 209458 b transmission spectra observations from JWST and HST using SANSAR, a newly developed planetary atmosphere modeling framework, with free, equilibrium chemistry and self-consistent grid retrievals. The overall best-fitting model with free retrievals ($χ^2_{\rm{red}}$=1.21) constrains its metallicity and C/O to be highly sub-solar, while equilibrium chemistry and grid retrievals ($χ^2_{\rm{red}}$=1.27 and 1.30, respectively) are consistent with solar values using STIS+WFC3+NIRCam observations. The retrieved abundances of H$_2$O and CO$_2$ are almost three orders of magnitude lower (highly sub-solar) with STIS+WFC3+NIRCam compared to just NIRCam, using free retrievals. NIRCam observations alone also result in misleading constraints on metallicity and C/O, with equilibrium chemistry and grid retrieval. We find that the model choice of varying C/H or O/H to vary the C/O in equilibrium chemistry retrievals leads to different metallicity constraints with NIRCam, but similar constraints with STIS+WFC3+NIRCam. We conclude that NIRCam observations alone can lead to overestimation of abundances for exoplanet atmospheres and, therefore, should be used in combination with UV/Optical and near-infrared observations to obtain robust constraints on abundances, C/O, and metallicity. In particular, even though we can detect the CO$_2$ feature with just NIRCam, we cannot constrain its abundances robustly without the optical baseline.

Figures (6)

• Figure 1: Ray diagram showing the calculation of path length traveled by a ray in an exoplanetary atmosphere at an arbitrary impact parameter $b$. Here $r^{i}_{\rm{up}}$ is the upper boundary radius of a layer, $d l_{\text{agg}}^i$ is the total path length traveled by the ray in the half geometry till the $i-1$ layer, and $d l^i$ is the path length of the ray in the $i^{\text{th}}$ layer.
• Figure 2: Figure showing line by line (LBL) transmission spectra comparison with opacity sampling at a resolution of 1000, 20000, and 100000 (all resolutions finally binned to R$\sim$100 for comparison) and with correlated-k at R$\sim$1000. Residuals compared with LBL are shown at the bottom, and the maximum error in each case is shown in the legend.
• Figure 3: Comparison between transmission spectra computed using the correlated-k opacity method and those from line-by-line (LBL) calculations. The LBL spectra have been binned to correlated-k wavelength grid (R$\sim$1000). Residuals are shown in the bottom panel.
• Figure 4: Comparison of SANSAR forward model with POSEIDON for a WASP-96 b type planet. Both SANSAR and POSEIDON use the POKAZATEL line list here. SANSAR samples lines from a 0.001 cm$^{-1}$ opacity file contrary to POSEIDON, which uses an opacity file with a resolution of 0.01 cm$^{-1}$.
• Figure 5: Comparison of forward models generated with SANSAR and ATMO for a WASP-96 b type planet. Both models were generated with the correlated-k method at a resolution of R$\sim$1000 and binned to R$\sim$200 for comparison.