Extended Data Fig. 7. A demonstration of how the redder wavelength coverage of NIRISS/SOSS drives the inference on cloud structure for WASP-39b.

We fit the NIRISS/SOSS spectrum (grey) using a suite of cloud models to derive the best-fit C/O ratio and metallicity. Here we demonstrate how the best-fit model for each cloud treatment changes as a function of what wavelength region we fit. a, The best-fit models when using the entire wavelength coverage of NIRISS/SOSS. b, The best-fit models when using λ < 2 μm, which excludes the overlapping region between orders on the detector. The reference spectrum (black) on both panels corresponds to the best-fit inhomogeneous droplet sedimentation model for the entire wavelength coverage. The fitted data are presented as dark-grey points. The quoted numbers in brackets in the legend are the respective χ²/N for each fit for the top (left value) and bottom (right value). The difference between cloud models is within the noise of the NIRISS/SOSS data when fitting to λ < 2 μm. It is clear that fitting the entire NIRISS/SOSS wavelength coverage results in a lower χ²/N and better fit. (https://github.com/afeinstein20/wasp39b_niriss_paper/blob/main/scripts/edfigure7.py).