On one night in August 2024, the James Webb Space Telescope recorded a faint point of mid-infrared light beside Alpha Centauri A. The source was more than 10,000 times fainter than the star and appeared about 1.5 arcseconds away, equivalent to a projected separation near two times the Earth-Sun distance.
There were good reasons to take it seriously. Tests made a passing asteroid, a distant background object and familiar image-processing artifacts unlikely. Yet Webb looked again in February and April 2025 and found no comparable point source. The possible planet, designated S1 in the analysis, had apparently vanished.
Two linked papers in The Astrophysical Journal Letters argue that a planet remains a plausible explanation. Their case combines one detection, two non-detections and millions of simulated orbits. This is one pair of studies, not settled confirmation of a planet.
One of the nearest targets is also one of the hardest
Alpha Centauri is the nearest stellar system to the Sun, at a little over four light-years. Its central pair contains Alpha Centauri A, a G-type star broadly similar to the Sun, and the somewhat smaller Alpha Centauri B. A third member, Proxima Centauri, lies much farther from the pair. An ESA/Hubble portrait of A and B makes them look cleanly separated, but a precision planet search has to contend with the detailed light pattern from both stars.
Webb’s Mid-Infrared Instrument, MIRI, used a coronagraphic mask to suppress Alpha Centauri A. That is only the beginning. Alpha Centauri B contributed its own off-axis glare, the stars moved rapidly against the background, and small pointing differences changed the residual patterns that survived subtraction.
The first of the two 2025 papers, led by Aniket Sanghi, describes the observations and image analysis. Depending on the processing method, S1 was detected at a signal-to-noise ratio between four and six, corresponding to a quoted significance of about 3.3 to 4.3 sigma. That is interesting, but it is not the sort of overwhelming single-epoch signal that makes follow-up optional.
The team therefore injected artificial point sources into the raw and processed images and asked whether the pipeline could recover them. It also tested whether S1 behaved like a fixed detector blemish, a residual copy of the stellar point-spread function or an unrelated moving object. Those checks support an astrophysical source, but no processing test can supply the second sighting needed to demonstrate an orbit.
The planet did not need to stop existing to disappear
A coronagraph does not leave a perfectly transparent field around a blacked-out star. Sensitivity falls near the central mask, while diffraction features, detector behavior and imperfect subtraction make some directions easier to search than others. An orbiting planet can move from a visible patch into a region where the same instrument and exposure no longer recover it.
That possibility became central after Webb’s February and April 2025 visits. Both observations were non-detections. The careful statement is not that the planet was seen and then ceased to exist. It is that S1 was seen once and would have been below the recovery threshold at the later positions for many otherwise acceptable orbits.
NASA’s August 2025 account of the work calls this a “disappearing planet,” but also states the decisive limitation: additional observations are required. The phrase describes an observing problem, not a confirmed world’s behavior.
Millions of orbits turned absence into a constraint
The second paper, led by Charles Beichman, combines the imaging limits with orbital and physical modeling. The team generated millions of possible paths, retained those consistent with S1’s August position and brightness, and removed orbits that would be dynamically unstable because of Alpha Centauri B.
The models also considered a point-like feature called C1 found in 2019 by the NEAR experiment on the European Southern Observatory’s Very Large Telescope. That earlier Nature Communications study treated C1 as a possible planet or a concentration of warm exozodiacal dust and explicitly required independent confirmation. S1 and C1 are not known to be the same object. The 2025 exercise asks what follows if they are.
Under that shared-object assumption, the accepted simulations gave a 52 percent probability that orbital motion would place the candidate in poor-sensitivity regions during both Webb follow-ups. In other words, missing it twice is not an especially improbable outcome. That calculation explains how the observations can fit a planet; it does not prove that a planet caused the original signal.
The surviving families generally have periods between two and three years. They favor an eccentricity around 0.4 and an orbit tilted by about 50 degrees, or the corresponding retrograde geometry, relative to the Alpha Centauri AB orbital plane. The star-planet distance would vary roughly between one and two astronomical units.
Saturn-mass is an estimate, not a measurement on a scale
S1 was measured in one MIRI filter centered near 15.5 micrometers. Turning that brightness into a planet requires assumptions about age, thermal evolution, radius, reflectivity and internal heat. The models point to a temperature around 225 kelvin, a radius roughly one to 1.1 times Jupiter’s and a mass between about 90 and 150 Earth masses.
Saturn contains about 95 Earth masses, which explains the useful shorthand “Saturn-mass.” The range extends well above Saturn, however, and no dynamical mass has been measured. A single infrared flux point also cannot yet provide an atmospheric composition.
The candidate would be unusual for direct imaging. Most directly imaged planets are young, hot giants far from their stars, where separation and leftover formation heat make them easier to distinguish. S1, if real, would be a mature, cooler giant much closer to its host star than the familiar directly imaged population.
SpaceDaily reported the initial Webb evidence in 2025. The enduring point is the same: this is the strongest imaging evidence yet for a planet around Alpha Centauri A, not a confirmed entry in the planet catalog.
A habitable-zone giant is not an Earth twin
The phrase “habitable zone” refers to the range of orbital distances where a suitably built rocky planet could maintain liquid water on its surface. NASA’s habitable-zone overview stresses that distance alone does not establish habitability. Atmosphere, pressure, composition, geology and stellar activity also matter.
For S1, the distinction is larger still. The candidate appears to be a gas giant with no solid surface like Earth’s. Its location would make it temperate by giant-planet standards, but it would not support surface life as we know it. Undetected moons could be discussed as possibilities, just as moons of giant planets are discussed in our own solar system, but Webb did not detect a moon, liquid water or any biosignature.
“Our nearest solar twin” also refers to the star, not the candidate planet. Alpha Centauri A resembles the Sun in spectral type and age more closely than the red dwarf Proxima Centauri does. The possible world beside it appears nothing like Earth on the evidence available.
Confirmation now has a moving target to find
The orbital simulations are useful because they turn three observing dates into predictions. Instead of searching blindly, future Webb visits can target times when a large fraction of viable paths place the source outside the coronagraph’s least sensitive region. A recovery at the right changing position would show common proper motion and begin narrowing the orbit.
Measurements in additional filters could test whether the source has the spectral energy distribution of a cooling giant planet or of warm dust. Repeated astrometry could link S1 to C1 or show that the apparent connection was coincidental. Radial-velocity and astrometric programs can add independent limits, although the binary stars make those measurements demanding too.
A clean non-detection at a time and place where nearly all viable planet orbits predict visibility would also be informative. It would push the interpretation back toward a transient artifact or some form of dust structure not captured by current tests.
For now, the disappearance is neither fatal to the planet case nor evidence for it by itself. Millions of simulations show how a Saturn-scale world could have slipped behind Webb’s observational blind spots. The next decisive step is simpler to state and harder to obtain: the point of light has to come back.