A new simulation of the habitable-zone planet TRAPPIST-1f arrives with three very different numbers: about 10 transits, about 50 transits, and more than 100.

Harper Rosenthal and colleagues asked how many observations by the James Webb Space Telescope might be needed to recover molecules from a hypothetical carbon-dioxide-rich atmosphere around TRAPPIST-1f. Their August 2026 preprint modeled both the planet and contamination from the surface of its ultra-cool red-dwarf star. In the simulation, roughly 10 transits produced strong evidence for carbon dioxide. Methane required about 50 transits to reach only weak evidence. Water could not be retrieved even after 100 simulated transits using both of the modeled JWST instruments. Rosenthal et al., 2026

Those are not detections. The researchers started by assuming a CO₂-rich, potentially habitable atmospheric model and asked whether JWST could recover its ingredients once the star was allowed to interfere. The exercise is interesting because the answer depends on a problem that sounds almost absurd at first: an exoplanet's atmosphere can be confused with its star.

The telescope never sees the planet by itself

Transmission spectroscopy works because a planet blocks a little of its star while crossing in front of it. At wavelengths where the planet's atmosphere absorbs more light, the planet appears very slightly larger. Measure the transit depth across many wavelengths and, in principle, molecules in the atmosphere leave recognizable patterns.

The phrase “the planet's spectrum” makes the measurement sound cleaner than it is.

The telescope first sees the combined light from the visible face of the star. During transit, it sees nearly the same star with one narrow region hidden behind the planet. The difference between those measurements contains information about the planet, but only if the hidden part of the star is spectrally representative of the star used as the baseline.

Often it is not.

Stars have spots, bright faculae, and temperature differences across their surfaces. Those regions have different spectra. If the planet crosses one sort of stellar surface while the rest of the visible star contains another, the subtraction can leave features that look planetary even though they originated on the star.

Benjamin Rackham, Dániel Apai, and Mark Giampapa formalized the problem for M-dwarf planets before JWST launched. They called it the transit light source effect: spectral differences between the portion of a star illuminating a transiting planet and the star's disk-integrated spectrum can be imprinted directly on the measured transmission spectrum. Their models showed that spots and faculae on M dwarfs could create false spectral features large enough to matter for atmospheric interpretation. Rackham, Apai, and Giampapa, 2018

That was a prediction about a measurement problem. JWST then ran into it.

TRAPPIST-1b showed how large the contamination can be

In 2023, Olivia Lim and colleagues used JWST's NIRISS instrument to obtain two transmission spectra of TRAPPIST-1b. They found moderate-to-strong evidence that unocculted stellar heterogeneities dominated both visits. One spectrum was consistent with unocculted starspots; the other showed signatures of unocculted faculae. Lim et al., 2023

The observation precision itself was excellent. The authors reported a combined precision of about 89 parts per million. The problem was that uncertainties from imperfect stellar models were roughly an order of magnitude larger.

Better instrumental precision does not solve a model error in the light source.

That distinction becomes particularly serious for the outer TRAPPIST-1 planets. A large hydrogen-rich atmosphere creates strong spectral features and is relatively easy to rule out. A secondary atmosphere around a small rocky planet can produce much subtler features. If the host star creates spectral structure larger than the atmospheric signal, collecting an exquisitely precise contaminated spectrum is not enough.

This is the history behind the new TRAPPIST-1f paper. Earlier forecasts for the system often assumed a much cleaner stellar surface. Early JWST observations showed that the convenient assumption was not good enough.

Ten transits for one molecule, fifty for another

Rosenthal and colleagues therefore put the contamination into the retrieval problem from the beginning.

They simulated a CO₂-rich atmosphere for TRAPPIST-1f and added what they describe as a “worst case” stellar-contamination spectrum. They then generated simulated observations for JWST's NIRSpec PRISM, covering roughly 0.6 to 5.3 micrometers, and MIRI LRS, covering roughly 5 to 15 micrometers. Their retrievals assumed that the spectra of the starspots themselves could be modeled accurately.

Carbon dioxide survived the exercise comparatively well. Around 10 transits gave the researchers strong Bayesian evidence for CO₂ in their modeled atmosphere.

Methane did not. It took around 50 transits to reach weak evidence.

Water was worse. Even with as many as 100 simulated transits using both instruments, the retrieval did not find evidence for H₂O. Rosenthal et al., 2026

The contrast matters more than any one number. A planet can contain several atmospheric molecules, observed by the same telescope around the same star, while the practical detectability of those molecules differs by an order of magnitude or more.

The cost is not only telescope hours. Each additional transit has to occur when the planet is actually crossing its star, and TRAPPIST-1f's orbital period is about nine days. A program requiring dozens of usable transits becomes a long scheduling commitment even before weather is removed from the equation, because JWST is in space. A hundred transit opportunities correspond to well over two years of orbital cycles.

That does not mean JWST would spend two continuous years watching TRAPPIST-1f. It means the geometry itself imposes a cadence. More photons cannot be ordered on demand.

There is another useful result in the paper: adding the modeled MIRI observations did not substantially improve the retrievals over NIRSpec alone. More wavelength coverage is not automatically more useful if the limiting problem lies elsewhere.

The star has become part of the atmospheric instrument

There is a tempting response to stellar contamination: model it away.

That is necessary, but the new paper also shows why it is not trivial. Rosenthal and colleagues assume accurate spectra for the spots used in their retrievals, while noting that reliable stellar-contamination models for ultra-cool M dwarfs remain a major challenge. If the stellar model is wrong, a retrieval can become impressively precise about the wrong decomposition of star and planet.

NASA has built a small mission around this problem.

Pandora began its science study in August 2026. The spacecraft is designed to observe exoplanets and their host stars together, with repeated measurements intended to help researchers separate changes in the star's spectrum from signals produced by a planet's atmosphere. NASA plans observations of at least 20 exoplanet systems. NASA, 2026

That is an unusual response to an atmospheric-measurement problem: spend more effort measuring the lamp.

For small planets around active cool stars, that may be what atmospheric spectroscopy requires. The spectrum arriving at the telescope is not stamped “star” and “planet” in separate layers. The instrument receives one stream of photons. The separation happens afterward, in models that have to know both objects well enough to decide which physical system produced each feature.

Pandora's job is to make one side of that subtraction less mysterious.

Sources

Ulix app

Curious Air

Curious Air turns an Android phone into an explorer for Wi-Fi, Bluetooth, cellular, GNSS, NFC, and device-sensor signals, with local inspection and optional timed logging.