Horizontal and vertical exoplanet thermal structure from a JWST spectroscopic eclipse map

TL;DR

This study uses JWST/NIRISS spectroscopic eclipse data of WASP-18b to perform true multidimensional eclipse mapping, resolving horizontal and vertical atmospheric structure across multiple wavelengths. By applying two independent mapping frameworks, Eigenspectra and ThERESA, the authors extract wavelength-resolved 2D maps and test 3D retrievals against HyDRA and Pyrat Bay, revealing a hotspot and a cooler ring with distinct thermal and chemical properties. The hotspot exhibits a thermal inversion and optical-opacity signatures, while the ring shows colder, more uncertain chemistry, with results broadly consistent with but sometimes challenging for existing GCMs that include magnetic drag and hydrogen dissociation effects. Overall, the work demonstrates the feasibility of 2D-3D atmospheric mapping with JWST, providing critical constraints on exoplanet dynamics, chemistry, and heat transport, and paving the way for similar analyses across a broader exoplanet sample.

Abstract

Highly-irradiated giant exoplanets known as "ultra-hot Jupiters" are anticipated to exhibit large variations of atmospheric temperature and chemistry as a function of longitude, latitude, and altitude. Previous observations have hinted at these variations, but the existing data have been fundamentally restricted to probing hemisphere-integrated spectra, thereby providing only coarse information on atmospheric gradients. Here we present a spectroscopic eclipse map of an extrasolar planet, resolving the atmosphere in multiple dimensions simultaneously. We analyze a secondary eclipse of the ultra-hot Jupiter WASP-18b observed with the NIRISS instrument on JWST. The mapping reveals weaker longitudinal temperature gradients than were predicted by theoretical models, indicating the importance of hydrogen dissociation and/or nightside clouds in shaping global thermal emission. Additionally, we identify two thermally distinct regions of the planet's atmosphere: a "hotspot" surrounding the substellar point and a "ring" near the dayside limbs. The hotspot region shows a strongly inverted thermal structure due to the presence of optical absorbers and a water abundance marginally lower than the hemispheric average, in accordance with theoretical predictions. The ring region shows colder temperatures and poorly constrained chemical abundances. Similar future analyses will reveal three-dimensional thermal, chemical, and dynamical properties of a broad range of exoplanet atmospheres.

Figures (15)

• Figure 1: $\,\vert\,$ Two-dimensional maps from the Eigenspectra method for each of the 25 spectroscopic bins. Colors indicate the temperature, while transparency indicates the relative contribution to the overall observed flux at the point of maximum visibility, based on the angle between a given point on the map and the line of sight to the observer. A maximum contribution of 1 indicates a latitude/longitude that is at the sub-observer point at some point during the observations. Dotted black curves delineate the three regions identified by the Eigenspectra mapping method. Evidence of multidimensional atmospheric structure can be seen in the varying hotspot temperature and shape with wavelength.
• Figure 1: $\,\vert\,$ Light curve fits from the Eigenspectra method for each of the 25 spectroscopic bins. Black points with error bars indicate wavelength and time-binned, systematics-corrected data from Coulombe et al. (2023)Coulombe2023, and red lines show best fits from the Eigenspectra method. The light curves are shown in planet flux ($F_p$) divided by stellar flux ($F_s$). The models fit the data well and show differences in brightness with wavelength, showing evidence of multidimensional atmospheric structure in the data.
• Figure 2: $\,\vert\,$ Retrieved longitudinal profiles at each wavelength range compared against GCMs. The Eigenspectra-retrieved profiles and GCMs including drag both show small hotspot offsets and sharp temperature gradients away from the substellar point at all wavelengths. The red lines and regions show the median retrieved longitudinal profiles and their 1$\sigma$ confidence intervals, respectively, measured with Eigenspectra for the twenty-five spectral bins considered. The profiles are obtained by weighting the retrieved 2D maps by the squared cosine of the latitude. The profiles are compared to two GCMs from ref. Coulombe2023 that matched the white-light map well - the SPARC/MITgcm (purple dash-dot line), which has uniform drag of timescale $\tau_{\text{drag}}$ = 10$^3$ s, and the RM-GCM (green dashed line), which includes a kinematic magnetohydrodynamical drag model with an internal magnetic field of $B \sim 20$ G. We note that the GCMs as shown here are processed to remove the "null space" of components which are physically inaccessible to eclipse mapping (see Methods and refs. Luger2021ChallenerRauscher2023ajNullSpace). Vertical dashed lines indicate zero longitude.
• Figure 2: $\,\vert\,$ Mean group (left), histograms of grouping across several MCMC samples (middle), and resulting eigenspectra (right) for Eigenspectra mapping fits using 2 (top), 3 (middle), and 4 (bottom) groups. For each set of plots, histograms are labelled by letters which are overplotted on the map in the latitude/longitude position from which they are drawn (positions were chosen both near and far from group edges). Groups 2 and 1 here correspond to the hotspot and ring groups discussed in the text. For 3 groups, the map shows a clear division between groups and all of the points show $\geq75$% of points assigned to a single group. For 2 groups, there is a similarly clear division, and notably the hotspot and outer groups have quite similar spectra to the corresponding groups in the 3-group case. For 4 groups, the mean group map does not show a clear division of groups, and the histograms show that the same point is sorted into different groups depending on the posterior draw. Additionally, the resulting spectra are not distinct (the group 0 spectrum is identical to the group 4 spectrum, which is why it is not visible on the plot). Therefore, we used 3 groups for this fit.
• Figure 3: $\,\vert\,$ Hotspot and ring group spectra from Eigenspectra bracket the full dayside-integrated spectrum. Black points with error bars (standard deviation; see Methods) show the Eigenspectra emission spectra from the hotspot and ring groups, while yellow points show the hemispherically-averaged homogeneous dayside emission spectrum Coulombe2023, binned in wavelength to match the group spectra. The spectra have been converted to brightness temperature by assuming blackbody emission for the planet in each bin and a PHOENIX emission model for the star. The HyDRA (purple) and Pyrat Bay (green) regions show 95.45% (2$\sigma$) credible regions from 1D atmospheric retrievals on each spectrum. Labels along the bottom show the wavelength ranges at which different atmospheric constituents create features in the spectrum. An average of the flux from the hotspot and ring regions produces a spectrum matching the dayside average (Extended Data Figure \ref{['fig:eigen_spectra_check']}). See the Methods for further discussion of the ring spectrum and the mismatch with associated retrievals.