Post

We observed the habitable zone planet TRAPPIST-1e with JWST to search for an atmosphere. You've seen the headlines, now let's dive into the science! 🧪 THREAD (1/N) #Exoplanets 🔭

Previously, on TRAPPIST-1: ➡️ No thick atmospheres on TRAPPIST-1b,c (Greene+2023, Lim+2023, Zieba+2023, Radica+2025, Gillon+2025). ➡️ Earth-like atmospheres ruled out for TRAPPIST-1d (Piaulet-Ghorayeb+2025) - see thread below. Now we turn to a planet more firmly in the habitable zone: TRAPPIST-1e.

TRAPPIST-1e is 92% Earth's size, 69% Earth's mass, and is illuminated by 66% of the integrated light that Earth receives. This means TRAPPIST-1e can potentially have liquid surface water *if* it has an atmosphere with a sufficient greenhouse effect. So TRAPPIST-1e was a priority target for JWST.

We observed TRAPPIST-1e four times with JWST in 2023 to measure how the apparent size of the planet changes with colour (i.e. transmission spectra) - more on why this took 2 years in a moment! Our spectra show *huge* wavelength-dependent features that are caused by active regions on the star ✴️

When we modelled the stellar contamination (similar to previous studies on TRAPPIST-1b,c, d), the models couldn't simultaneously explain the entire wavelength range. Simply put, our stellar models for ultra-cool M-dwarf stars like TRAPPIST-1 don't work 😱 So we had to try something new...

We turned to Gaussian Processes (GPs) to fit the stellar contamination affecting the TRAPPIST-1e spectra. Since: Observed_spectrum_i = contamination_i * planet_spectrum The idea is to extract the time-independent (non-GP) common factor caused by any planetary atmosphere.

Using GPs to account for the stellar contamination, we combined the time-independent spectral information from the four transits to produce the spectrum of TRAPPIST-1e shown in the press release. We then turned to atmospheric models to see if there were any signatures of atmospheric absorption.

Our first result was a firm rejection of any significant amount of hydrogen in TRAPPIST-1e's atmosphere. Irrespective of the cloud-surface pressure, we find a H2 abundance limit of < 80% (to 3σ). This is a significant improvement over what was possible with Hubble data.

The observations, stellar contamination GP magic 🪄, and H2-upper limit we've discussed so far are covered in our first TRAPPIST-1e paper, led by Néstor Espinoza at STScI (not on Bluesky). Be sure to check out the paper! iopscience.iop.org/article/10.3... Next, we looked for secondary atmospheres.

In Paper #2, we ran a grid of atmospheric models considering combinations of strong infrared absorbers (CO2 / CH4) and transparent background gases (N2 / H2). The figure below (from Glidden+2025) shows the range of excluded partial pressures. Big takeaway: large CO2 concentrations are unlikely.

Intriguingly, forward models with N2 + CH4 provided a great fit to TRAPPIST-1e's transmission spectrum 😯 We found the same solution independently through atmospheric retrievals, which latched onto CH4 absorption as a potential explanation. 🔍 But this is not (yet!) an atmospheric detection.

Statistically, our current four-transit spectrum of TRAPPIST-1e can also be fit by a flat line (i.e. a featureless spectrum). So we can't rule out a bare rock with these data. There's also the important caveat that an incomplete stellar contamination correction could also imprint spectral features.

Technical point: retrievals of flat transmission spectra for rocky planets usually result in corner plots resembling the prior. For TRAPPIST-1e, we don't see this behaviour, with the CH4 posterior pushing to include this molecule. We haven't detected CH4, but future observations can assess this.

Our constraints on potential atmospheres with molecules heavier than H2 and He (secondary atmospheres) are presented in our second TRAPPIST-1e paper, led by @ana-glidden.bsky.social at MIT. Be sure to check out the paper! iopscience.iop.org/article/10.3... So what comes next?