No thick carbon dioxide atmosphere on the rocky exoplanet TRAPPIST-1 c

The study seeks to determine whether the terrestrial exoplanet TRAPPIST-1 c possesses a substantial atmosphere and, if so, to constrain its composition and surface pressure using direct measurements of thermal emission. Atmospheric properties of rocky exoplanets are poorly known and depend on volatile inventories, outgassing, escape, and potential atmospheric collapse. Planets orbiting M dwarfs may be especially vulnerable to atmospheric loss during the star’s long pre-main-sequence phase. Previous thermal emission measurements for similar small, hot rocky planets (LHS 3844 b, GJ 1252 b, and TRAPPIST-1 b) indicated high dayside temperatures with little to no heat redistribution and no detectable CO2 absorption, motivating observations of cooler targets like TRAPPIST-1 c that may have retained atmospheres. This work uses JWST/MIRI to measure the 15 µm dayside emission of TRAPPIST-1 c, directly probing CO2 absorption and atmospheric heat redistribution.

Prior work has shown that several hot terrestrial exoplanets likely lack thick atmospheres: LHS 3844 b and GJ 1252 b exhibit dayside temperatures consistent with no heat redistribution and no CO2 absorption; JWST measurements for TRAPPIST-1 b similarly indicate a bare rock. Theoretical studies suggest atmospheres of rocky M-dwarf planets evolve under strong XUV irradiation, potentially leading to water loss and abiotic O2 build-up, with CO2 from outgassing possibly accumulating to high pressures. Observational strategies for temperate Earth-sized planets with JWST and modeling frameworks for identifying candidate atmospheres via eclipse photometry support focusing on CO2 features near 15 µm. Interior-structure modeling of TRAPPIST-1 planets provides independent pressure constraints, and Venus-like atmospheric analogues offer a comparative framework given TRAPPIST-1 c’s insolation being only ~8% higher than Venus.

Observations: Four secondary eclipses of TRAPPIST-1 c were observed with JWST/MIRI imaging (General Observer program 2304) using the F1500W filter (3-µm-wide band centered at 15 µm) on 27 Oct, 30 Oct, 6 Nov, and 30 Nov 2022. Each visit lasted ~192 minutes, covering the ~42-minute eclipse and out-of-eclipse baseline. The FULL subarray was used, yielding 1,190 integrations total. Data reduction: Four independent reductions were performed using the Eureka! pipeline and custom software. Aperture photometry extracted stellar light curves; instrumental systematics were modeled with polynomials in time, exponential ramps, and decorrelation against PSF position/width. Residual scatter RMS across analyses was 938–1,079 ppm, within 1.06–1.22× the photon noise (with corrected gain). Eclipse depths were estimated via MCMC fits marginalizing over systematics and astrophysical parameters. The four reductions agreed within 1σ; the final eclipse depth adopted the mean and uncertainty across reductions, with an added 6 ppm in quadrature to account for reduction-dependent systematics, yielding f/f* = 421 ± 94 ppm. Reduction specifics:

• SZ: Started from raw uncal files; JWST Calibration Pipeline v1.8.2 via Eureka! stages 1–3. Custom ramp-jump thresholds (visits 1–4: 70, 60, 70, 50). Skipped photom conversion to keep DN s−1. Masked DO NOT USE pixels; measured centroid and PSF widths via 2D Gaussian; chosen target aperture radius 4 px; background annulus 25–41 px. Removed first 10 integrations per visit to avoid initial MIRI ramp; iterative 5σ clipping found no outliers; final integrations per visit: 288, 287, 287, 288.
• ED: Eureka! stages 1–5; ramp-fitting weights set to uniform; photom step off; defined subarray region; masked and interpolated bad pixels; aperture photometry with background annulus 20–35 px; recorded PSF centers/widths; sigma-clipped 40 outliers with a 10-integration boxcar; optimized aperture radii 3.7, 4.0, 3.6, 3.8 px per visit.
• MG: Eureka! stages 1–2 with uniform ramp weights and jump correction disabled; subsequent processing in IRAF/Fortran: unit conversion to electrons, 2D Gaussian centroid and FWHM, circular/annular apertures via IRAF/DAOPHOT; normalized light curves and 5σ clipping with 20-min moving median; per-visit aperture optimization; photometric uncertainties included star and background photon noise, readout, and dark noise (gain 3.1 e− ADU−1).
• PT: Analysis of level-2 calints products from MAST; 2D Gaussian centroid/FWHM; fixed- and variable-aperture photometry (variable radii = c × smoothed FWHM, c scanned); additional details in Methods. Modeling and interpretation: The measured eclipse depth was translated to a 15 µm brightness temperature using stellar/planet parameters, yielding 380 ± 31 K. Atmospheric model grids (cloud-free O2-dominated with CO2 ranging 1–10,000 ppm; pure CO2) with surface pressures 0.01–100 bar and Bond albedo 0.1 included both radiative absorption and heat redistribution. Comparisons also included forward models of Venus-like atmospheres (96.5% CO2, 3.5% N2 with/without H2SO4 aerosols, 10 bar) from a coupled climate–photochemistry model, and bare-rock emission models for multiple surface compositions with/without space weathering. Interior-structure constraints from prior work contextualized allowable surface pressures. Atmospheric evolution models explored initial water inventories (0.1–100 Earth oceans) and stellar XUV saturation fractions (10−4–10−1), tracking outgassing and escape to infer final O2 surface pressure after 7.5 Gyr.

• Measured secondary eclipse depth at 15 µm: f/f* = 421 ± 94 ppm.
• Inferred dayside brightness temperature: 380 ± 31 K.
• Dayside temperature lies between zero-redistribution (430 K) and full-redistribution (340 K) limits for a zero-albedo planet, consistent with either moderate heat redistribution (ε = 0.66+0.23−0.39) or a non-zero Bond albedo for a rocky surface (A_B = 0.57+0.37−0.25).
• Atmospheric constraints from model grid comparisons: • All thick atmospheres with P_surf ≥ 100 bar ruled out. • Assuming CO2 ≥ 10 ppm, P_surf ≥ 10 bar ruled out. • For cloud-free, pure CO2 atmospheres, P_surf ≥ 0.1 bar ruled out.
• Venus-analogue atmospheres (10 bar, 96.5% CO2, 3.5% N2): • Cloud-free Venus-like model disfavoured at 3.0σ. • Cloudy (H2SO4 aerosols) Venus-like model disfavoured at 2.6σ; aerosols warm the 15-µm core slightly, marginally improving consistency.
• Bare-rock models (basaltic, feldspathic, Fe-oxidized, granitoid, metal-rich, ultramafic) with space weathering considered: all are consistent with the measured eclipse depth; highest-albedo unweathered feldspathic/granitoid surfaces are marginally worse (consistent at ~2σ).
• Interior-structure models independently cap P_surf to ≤160 bar (3σ) and ≤80 bar (1σ), consistent with observed constraints.
• Formation and volatile inventory implications: For a typical M-dwarf XUV saturation fraction (L_XUV/L_bol)_sat ≈ 10−3, final O2 surface pressures of 0.1–100 bar are possible depending on initial water. The measured eclipse depth disfavors very high surface pressures and implies initial water abundance <9.5 ± 3 Earth oceans; for CO2 >10 ppm, <4.0 ± 1 Earth oceans. Overall, TRAPPIST-1 c likely formed volatile-poor and lacks a thick CO2-rich atmosphere.

The 15 µm dayside emission detection directly addresses the presence and nature of TRAPPIST-1 c’s atmosphere. The high brightness temperature and lack of a pronounced CO2 absorption-induced flux suppression in the MIRI F1500W band are incompatible with thick CO2-rich atmospheres across a broad range of surface pressures, favoring either a thin atmosphere with limited heat redistribution or a bare-rock surface. The intermediate dayside temperature relative to the no- and full-redistribution limits suggests either moderate atmospheric heat transport or a reflective rocky surface. Venus-like compositions are statistically disfavored, indicating that despite similar stellar irradiation to Venus, TRAPPIST-1 c’s atmospheric evolution diverged, likely due to volatile loss and limited outgassing. Bare-rock spectral models, especially for fresh low-albedo or weathered surfaces, match the observed eclipse depth. Linking these atmospheric constraints with evolutionary models implies a low initial water inventory and/or efficient escape, consistent with expectations for planets around active M dwarfs. This has broader implications for the volatile budgets and potential habitability of other TRAPPIST-1 planets, motivating multi-wavelength thermal measurements and transmission spectroscopy to assess atmospheres across the system.

Using JWST/MIRI eclipse photometry at 15 µm, the study measures the dayside emission of TRAPPIST-1 c and finds a planet-to-star flux ratio of 421 ± 94 ppm, implying a dayside brightness temperature of 380 ± 31 K. Comparisons with atmospheric and surface models rule out thick, CO2-rich atmospheres over wide pressure ranges and disfavor Venus-like scenarios, while thin atmospheres or bare-rock surfaces remain consistent. Coupled with atmospheric evolution modeling, the results suggest a volatile-poor formation history with initial water inventories below ~10 Earth oceans (and below ~4 Earth oceans for CO2 >10 ppm). These findings indicate that rocky planets around M dwarfs, including those in TRAPPIST-1, may generally possess limited volatile reservoirs. Future work should target multi-band thermal emission and spectroscopy for TRAPPIST-1 c and its sibling planets to refine atmospheric composition, heat redistribution, albedo, and volatile inventory inferences.

• Single-band eclipse measurement (MIRI F1500W) limits compositional discrimination and vertical temperature profile constraints; conclusions rely on model comparisons centered on the 15 µm CO2 band.
• Atmospheric model grid assumes specific Bond albedo (0.1) and cloud-free conditions for O2/CO2 mixtures; different albedos/cloud properties could shift constraints.
• Venus-analogue comparisons are disfavored at moderate significance (2.6–3.0σ), not definitive exclusions.
• The analysis provides dayside-averaged constraints; nightside temperatures and global heat redistribution remain indirectly inferred.
• Instrumental systematics are mitigated with multiple independent reductions, but a small systematic term (6 ppm) is added to the uncertainty; unmodeled systematics could persist.
• Interior-structure and evolution inferences depend on stellar XUV histories and escape efficiencies that carry uncertainties.