For years, 55 Cancri e looked like a test of how thoroughly a nearby star could strip a rocky planet. It completes an orbit every 17.7 hours, close enough for the surface to be molten, and its original hydrogen and helium envelope was expected to have escaped long ago.

Update. The study’s lead author, Ignas Snellen of Leiden Observatory, has said that a June analysis reporting variable carbon monoxide emission from 55 Cancri e is wrong. In a public post he wrote that “the pre- and post-eclipse structures are nearly vertical, accompanied by negative components, and almost exact mirror images. This is the characteristic signature of a small temporal wavelength drift,” that “After correcting for these drifts, no significant CO signal remains,” and that “We have therefore withdrawn our Nature Astronomy manuscript.” The arXiv abstract page still lists the 10 June 2026 submission and does not say the preprint was withdrawn.

A June analysis of James Webb Space Telescope observations reported carbon monoxide high in the atmosphere, changing between observations, with atmospheric models favouring a hydrogen-rich composition linked to a reduced magma ocean. This was one study, not settled consensus. It is a preprint that had been submitted to Nature Astronomy, not a peer-reviewed result.

The preprint described a measurement made by waiting for the planet to disappear

Ignas Snellen of Leiden Observatory and colleagues reanalysed five Webb NIRCam observations from November 2022 and April 2023. Each visit covered a secondary eclipse, when the planet passed behind its star. By comparing the star-only signal during eclipse with the combined light before and after it, the team isolated the dayside’s faint infrared contribution.

Earlier analyses smoothed the spectra into broad wavelength bins. Snellen’s team retained NIRCam’s native resolving power across 3.9 to 5 micrometres and cross-correlated the data with the pattern expected from individual molecular lines. The preprint reported an approximately eight-sigma carbon monoxide emission signal in one visit, possible three-sigma signals in two others and no significant signal in the other two. Snellen has since said that after the correction “no significant CO signal remains.”

The preprint reported that the strongest feature appeared near the planet’s expected orbital speed and was absent while the star hid the planet, and it treated those checks as support for a planetary origin. Snellen has since written that “By an unfortunate coincidence, in Visit 1 these line up almost exactly with the radial-velocity trail expected for planetary CO emission,” and that “After correcting for these drifts, no significant CO signal remains.”

Hydrogen is a modelled explanation, not a direct detection

The preprint said the strongest signal was difficult to reconcile with a hydrostatic atmosphere. In its modelling, that signal required a steep thermal inversion near pressures of about 1 to 10 millibars and a carbon dioxide abundance at least three orders of magnitude lower than carbon monoxide, which would otherwise mask the carbon monoxide signal.

Across 8,136 self-consistent atmospheric models, the preprint found that hydrogen-rich cases tended to produce the steepest inversions and the highest ratios of carbon monoxide to carbon dioxide. The authors interpreted that preference as consistent with outgassing from a reduced magma ocean, an interior of relatively low oxygen fugacity. Snellen has since said the referees brought to light “an issue with the data and therefore with our interpretation.”

The preprint’s abstract presents a hydrogen-rich atmosphere as what its modelling found most readily, not as a direct detection of molecular hydrogen. Hydrogen richness and a link to gas leaving a molten interior were interpretations of the carbon monoxide signal. Snellen has since said that after the correction, “no significant CO signal remains.”

What the preprint said about changes between visits

The preprint treated the variation across five visits as a sign that the carbon monoxide signal might be intermittent rather than a persistent feature of a static atmosphere. It considered changes in temperature structure, time-dependent outgassing, clouds and atmospheric escape. It described a feedback cycle in which material released from the magma ocean forms clouds that cool the surface and temporarily suppress further outgassing until the clouds dissipate.

The preprint also said variability in the magma ocean, for example through tidally driven flows, could alter local surface temperatures and outgassing rates. It said the observations are most naturally interpreted as probing denser, bound atmospheric layers. The abstract says the variable signal may be linked to variable atmospheric outflow. The preprint said carbon monoxide is comparatively stable, yet maintaining large abundances in the low-density exosphere may require shielding and continuous replenishment, and that in the unshielded ultraviolet field carbon monoxide is expected to survive for at most tens of minutes.

The preprint said the data did not allow the authors to distinguish these possibilities. It also reported instrumental variations at about the 100 to 200 parts per million level in the light curves, and said the cross-correlation analysis was insensitive to those broad, wavelength-independent effects. Snellen has since written that the other visits “have smaller and differently oriented drifts, explaining their absent or marginal signals,” and that “After correcting for these drifts, no significant CO signal remains.”

The bare-rock expectation was never the only one

Astronomers had good reason to think 55 Cancri e lost its primordial atmosphere, but that was not the same as agreement that the planet had no atmosphere worth studying. Earlier Spitzer measurements produced competing clues about the planet’s heat distribution, surface and possible blanket of gas.

In 2024, a team led by Renyu Hu of NASA’s Jet Propulsion Laboratory reported in Nature that Webb observations from four to 12 micrometres favoured a substantial secondary atmosphere rich in carbon monoxide or carbon dioxide over a bare surface carrying only vaporised rock. The paper, summarised by NASA, also proposed that dissolved gases could bubble out of a magma ocean and continually renew the atmosphere.

Uncertainty about composition persisted. A 2025 Astronomy & Astrophysics study, with M. Zilinskas as first author and L. J. Janssen among the co-authors, tested more than 25,000 models and statistically favoured a hydrogen-free, nitrogen-dominated atmosphere, while finding that several alternatives remained viable. The June preprint examined the same five NIRCam eclipses at their native spectral resolution and reported a different preference, for a hydrogen-rich atmosphere. Snellen has since said that result is wrong.

Repeated measurements of the planet remain possible

NASA’s catalogue puts 55 Cancri e at about 1.9 times Earth’s radius and eight times its mass. At roughly 41 light-years away, orbiting a bright star, it is unusually accessible for repeated measurements of a rocky exoplanet. Its short year also gives observers frequent eclipses.

Snellen wrote that the team plans “to report the wavelength-stability analysis and injection tests in a separate paper.”

The preprint’s picture of variable carbon monoxide and a hydrogen-rich atmosphere is the result Snellen has said is wrong. The 2024 Nature study of a volatile atmosphere on 55 Cancri e is a different paper.