Direct Spectroscopy of 51 Eridani b with JWST NIRSpec

TL;DR

This study demonstrates direct spectral detection of the exoplanet 51 Eridani b with JWST/NIRSpec in fixed-slit mode, achieving a 4.8σ signal at 3-5 μm without a coronagraph by cross-correlating continuum-subtracted spectra with high-resolution atmospheric templates. The robust, multi-faceted analysis detects CH$_4$ and CO as the primary absorbers, providing the first direct evidence of these molecules coexisting in chemical disequilibrium in a directly imaged planet. A comprehensive atmospheric-model fit to space- and ground-based data yields $T_ ext{eff} ≈800$ K, $ ext{log} ext{g}≈3.75$, $[ ext{M/H}]=0.7$, $C/O≈0.458$, and other parameters with an errorbar inflation factor $\hat{e}≈1.74$, highlighting residual systematic uncertainties in spline-based continuum subtraction. The work underscores both the potential and current limitations of fixed-slit high-contrast spectroscopy for exoplanets, suggesting improvements in covariance modeling and advocating a preference for IFU observations for future campaigns while outlining strategies for follow-up measurements of CO$_2$ and other tracers.

Abstract

We present high-contrast direct spectroscopy of the low-mass, cool exoplanet 51 Eridani b (2-4 M$_\textrm{Jup}$, $\sim$750 K) using JWST / NIRSpec in a fixed-slit configuration (F290LP / G395H, $3-5\,μ$m, R$\sim$2,700). A cross correlation analysis between the continuum-subtracted data and atmospheric forward models indicates a detection of molecular signals of planetary origin at $4.8σ$ at the expected position and velocity of the planet. The detection of the planetary signal is driven primarily by molecular features from methane and carbon monoxide, providing the first direct confirmation of these two molecules coexisting in chemical disequilibrium in the atmosphere of 51 Eridani b. A new comprehensive atmospheric model analysis shows consistency between the ground-based IFU spectroscopy and the NIRSpec data, with the best-fit model parameters: $T_\mathrm{eff}$ = 800$^{+21.5}_{-55.5}$ K, $\log g$ = 3.75$^{+0.09}_{-0.37}$, $[\mathrm{M}/\mathrm{H}]$ = 0.7$^{+0.07}_{-0.21}$, $\textrm{C}/\textrm{O}$ = 0.458$^{+0.08}_{-0.09}$, $\log K_\mathrm{zz}$ = 3$^{+0.47}_{-0.73}$, $R_\mathrm{P}$ = 1.36$^{+0.07}_{-0.03}$ $R_\mathrm{Jup}$, $f_\mathrm{hole}$ = 0.3$^{+0.10}_{-0.07}$, and the NIRSpec errorbar inflation parameter: $\hat{e}$ = 1.74$^{+0.02}_{-0.03}$. We conclude with a discussion on the lessons learned between the fixed slit and IFU-based high contrast spectroscopic methods from our observing program, including some possibilities to improve the analysis method.

Figures (11)

• Figure 1: Diagnostic plots generated during data reduction. (Left) Detection significance as a function of the node spacing hyperparameter for each individual data sequence showcasing the robustness of the detection, as well as the optimal node spacing. (Right) Best fit coordinates in slit A1 (top) and slit A2 (bottom) and corresponding regularized polynomial fit as a function of wavelength computed during the centroiding / coordinate correction step.
• Figure 2: Detection of the planet 51 Eridani b from moderate resolution spectroscopic observations with JWST/NIRSpec. Top: Visualization of the point cloud in sky coordinates (left) and IFU-aligned coordinates (center). In the background at half opacity is the best fit synthetic WebbPSF from the coordinate correction / centroiding step. In the foreground at full opacity is an interpolation of the continuum-preserving point-cloud for the combined planet and speckle side observations. The "slit" appears $\sim$0.05 arcsec wide in this figure because that is the scale subtended by the sub-pixel dither pattern, not the full 0.2 arcsec width of the physical slit aperture. (Right) A demonstration of the offset in IFU $Y$ between the peak intensity of the stellar PSF and the location of the peak of the cross correlation function, indicating the signal is not a result of amplified stellar noise. Bottom: Cross-correlation based detection maps in the IFU $Y$ and planetary radial velocity phase space. (Left) Joint detection map for all planet side datasets, (Right) joint detection map for all speckle side datasets. The white dashed lines indicate the expected position and velocity of the target.
• Figure 3: (Upper) Template spectra used for computing the CCF shown with perturbations from removing individual molecules in the radiative transfer calculation. (Middle) Molecular contribution functions calculated using picaso.jdi.get_contribution. The curve shows the pressure layer where the optical depth per species is approximately unity. (Lower) A visualization where the absorption features are shown per species using colored banding with variable opacity to denote deeper absorption.
• Figure 4: Searching for individual molecules in the atmosphere of 51 Eridani b with "leave one molecule out" cross correlations. (Top) Cross correlation functions computed on all four templates showing the effect of removing certain molecules from the atmospheric model across the entire wavelength range. (Middle) Restricted wavelength test in the region of CH4 dominance. (Bottom) Restricted wavelength test in the region of CO dominance.
• Figure 5: (Upper) Four primary parameters of the Diamondback evolutionary model grid from left to right: Luminosity, Effective Temperature, Surface gravity, and Radius as a function of age (x-axis) and mass (color). Mass ranges from 0.0005 to 0.08 M$_\odot$ going from dark blue to gold or equivalently 0.524 Mjup to 84 Mjup. (Lower) On the left, the assumption on Mass and Age for the prior distribution, and on the right the interpolation of that distribution onto the four output parameters of the distribution, which forms the prior on effective temperature, surface gravity, and radius. The prior is assumed flat over the entire grid on all of the other atmospheric parameters.