On 28 November 2025, the James Webb Space Telescope was studying a planet astronomers already knew. Its spectrograph found another one.
Five days later, a different team pointed the Very Large Telescope at the same star for another study. A faint point appeared on the other side of the system. It was the same new world.
The teams had found Beta Pictoris d independently, then deliberately kept their analyses apart. The overlap was not a simple repeat. Webb identified a chemical pattern and a line-of-sight speed inside the glare of a debris disk. The VLT caught a moving point, then gave old observations a new target. Images from six epochs showed that point travelling with the star over 11 years.
One planet had left two kinds of receipt.
That is why this discovery is more revealing than the usual announcement of a dot beside a star. It shows what “found” means when the object is roughly 63 light-years away, hidden in processed infrared data and too light for astronomers to have weighed through its pull on the star. Some properties were measured. Others were inferred. The difference is the story.
Webb searched the spectrum, not just the picture
Beta Pictoris is bright, young and surrounded by an edge-on disk of dust and gas. That makes the system scientifically rich and visually awkward. The disk puts extra light exactly where an outer planet is likely to appear.
Webb's NIRSpec instrument approached that problem as thousands of tiny spectra. Its integral-field unit records a position on the sky and a spectrum for every element in the field. The observing programme was designed to study Beta Pictoris b. When the team searched the resulting data cube, it found a localised pattern southwest of the star.
Dust reflecting starlight should have produced a relatively smooth continuum. The unexpected source contained regularly spaced carbon-monoxide absorption lines between about 4.4 and 5 micrometres. Methane appeared near 3.3 micrometres. Follow-up with Webb's MIRI instrument added water absorption at longer wavelengths.
The brightness alone proved little. Bright blobs can be processing artefacts or structures in the disk. Molecular lines are harder for those alternatives to imitate. The spectroscopy study reports that a dust clump or background star could not make the observed combination. A brown dwarf in the same line of sight remained a physical possibility, but an extraordinarily unlikely one.
The spectrum supplied motion as well as chemistry. The Doppler shift of the carbon-monoxide lines put the object's radial velocity at roughly 10 to 16 kilometres per second in the two Webb epochs. Beta Pictoris b, visible on the other side of the star, was moving the other way along our line of sight. The new source's position and velocity fit a planet orbiting in the disk plane.
The team estimated the chance of an unrelated brown dwarf landing within two arcseconds of the star and ten degrees of the disk plane at less than one in ten million—even before using the velocity. The paper therefore treats the first NIRSpec observation as sufficient for a high-confidence discovery. The March 2026 NIRSpec and MIRI observations strengthened and characterised it; they did not rescue an unconfirmed blob.
The distinction changes the double discovery. These were not two broken halves that became a planet only when joined. They were independent discoveries with different blind spots.
The VLT made old light move
The second discovery began on 3 December 2025 with ERIS, an infrared instrument on one of ESO's 8.2-metre Very Large Telescope units in Chile. The team was also studying Beta Pictoris b. It did not use a coronagraph for this observation. After modelling and subtracting the star's pattern, a fainter point remained just over one arcsecond to the southwest, detected at a signal-to-noise ratio of 14.9.
A point in one heavily processed image is a candidate. A sequence of positions can become an orbit.
The researchers searched observations made for earlier projects. They recovered the source in Webb/NIRCam images from 2023 and March 2025, and in VLT/SPHERE data from 2019 and 2020. The hardest recovery came from December 2014. In that frame the new planet lay only about 45 milliarcseconds from Beta Pictoris b, which was roughly 100 times brighter in the relevant bands. The team modelled and removed b before the fainter source emerged from its wings.
Those frames did not change after 2014. The new detection changed the question researchers could ask of them: if this point is one orbiting object, where should it have been at each earlier date?
Across the six epochs, the measured positions diverged from the path expected for a stationary background source by more than 24 standard deviations. A sufficiently sensitive 2011 observation showed no object where that background interloper should have been. The source moved outward along the same nearly edge-on plane as the disk and the two known planets.
The imaging and orbit study did more than find a second dot. It turned archival light into an 11-year motion baseline.
What the instruments actually proved
The discoveries overlap, but their strongest evidence is not interchangeable.
| Evidence | What was measured | What it supports | What remains model-dependent |
|---|---|---|---|
| NIRSpec and MIRI spectra | Carbon monoxide, methane and water features; radial velocity | A gaseous object whose motion and position fit an orbit around Beta Pictoris | Temperature and mass from continuum-subtracted line shapes |
| ERIS detection | Infrared contrast and a point-source position in December 2025 | A strong ground-based detection at the expected location | Physical brightness after calibration and processing |
| Six archival/current image epochs | Relative position from 2014 to 2025 | Common motion, a bound companion and initial orbit | Semimajor axis and eccentricity depend partly on priors and stability choices |
| Infrared colours and luminosity | Brightness in several filters | A cold atmosphere with probable carbon-dioxide absorption | Metal enrichment, radius, temperature and mass require atmosphere/evolution models |
The distinction is easiest to see in the mass. The imaging team estimated 2.4 ± 0.6 times Jupiter's mass. That is not a dynamical weighing like the measurements available for Beta Pictoris b and c. It starts with the new planet's luminosity, adopts a system age of 23 ± 8 million years, borrows a bolometric correction from the similar planet 51 Eridani b, then compares the result with ATMO hot-start evolutionary models.
The Webb team arrived at a broader two-to-four-Jupiter range. Its continuum-subtraction technique was excellent for exposing narrow molecular features but removed the continuum that normally anchors temperature. Different atmosphere grids fitted line shapes at temperatures as high as 900–1,200 kelvin only by implying implausibly small radii. The authors judged about 600–800 kelvin more consistent with evolutionary radii, prior non-detections and plausible chemistry. That chain is informative, but it is not a thermometer placed in the atmosphere.
The imaging analysis gives 600 kelvin, with a range of 540 to 645, and a modelled radius of 1.26 Jupiter radii. The agreement between two analyses is encouraging. It does not turn their shared assumptions into direct measurements.
Two orbital answers can both be honest
The orbit offers a second lesson in how numbers acquire authority.
With 11 years of positions, the imaging team tightly constrained the orientation. Beta Pictoris d's orbit is inclined about 89 degrees to our view, closely aligned with planets b and c. The path is almost edge-on.
Other parameters were looser. The raw orbital posterior included several families: moderate-distance, low-eccentricity paths, plus more eccentric paths caught near their nearest or farthest points. About 28% of the fitted solutions crossed the orbit of Beta Pictoris b and were likely unstable. After removing those, the team reported a semimajor axis of 26.0 astronomical units, with a one-standard-deviation range from 19.9 to 28.2, and a period of 91 years, with a range from 64 to 109.
The Webb analysis had only two astrometric and velocity epochs for d. Its own fit centred near 24 astronomical units but was broad. When the team simulated the system for five million years, stable arrangements preferred a semimajor axis above 30 astronomical units.
“Twenty-six” and “more than 30” look like a contradiction only if the method disappears. One is a joint astrometric posterior filtered to exclude orbit crossings. The other is a preference added by a particular long-term stability simulation to a short-baseline fit. Future observations will move the planet along its 90-ish-year path and test both.
For now, the exact honest answer is a distribution, not a single ring drawn around the star.
The faintest record comes with a footnote
ESO called Beta Pictoris d the faintest exoplanet directly imaged from Earth. The organisation also supplied the sentence that keeps the superlative true.
The record uses absolute magnitude: brightness corrected for the system's distance. It ranks the planet's intrinsic faintness, linked to its size and temperature. It does not mean no telescope has recorded an apparently dimmer planet farther away.
The “100 times fainter” comparison is similarly specific. In the 2014 SPHERE bands, Beta Pictoris b was about 100 times brighter than d, and their projected positions nearly overlapped. That explains how a real source could sit in a famous dataset for years without being recognised. It does not mean every instrument at every wavelength sees the same ratio.
The result still pushes ground-based imaging into unusual territory. Most directly imaged planets are young and massive enough to glow strongly in infrared light. Beta Pictoris d is also young, but models place it at only a few Jupiter masses and about 600 kelvin. ERIS found it at 3.79 micrometres, near a favourable part of a cold giant's thermal output. A large mirror and a short optical path helped on the ground; stable space spectroscopy attacked the disk glare from a different direction.
The VLT and Webb did not need to become one telescope. Their errors were different.
A planet at the edge of an unfinished system
Beta Pictoris is only the second known system with at least three directly imaged planets; HR 8799 is the other. Its youth makes the planets warm enough to study, while its edge-on debris disk records a system that is still dynamically busy.
Planet d sits outside b and c, near the disk's uncertain inner boundary. Models had already suggested that another planet beyond b might be needed because the two known giants could not clear material out to the observed edge. Using a disk edge of 40 ± 5 astronomical units, the imaging team calculated that a single sculpting planet should have at least 1.3 Jupiter masses and orbit no farther out than about 31 ± 4 astronomical units. Their adopted values for d fall inside that region.
That is compatibility, not a historical photograph of the clearing. The disk is viewed edge-on, so its inner edge may lie anywhere from roughly 30 to 50 astronomical units. The analysis cannot rule out more planets. The new world might also scatter comets inward, but that is another dynamical possibility awaiting a firmer mass and orbit.
Its atmosphere holds a separate clue. The imaging colours suggest strong carbon-dioxide absorption and possibly more metals than free-floating objects of similar age and temperature. Current atmosphere models do not fit every band equally well. The authors call for spectroscopy and, eventually, a dynamical mass.
The discovery closes the easy question—there is a third planet—and opens the harder ones.
The archive was part of the instrument
Direct images feel decisive because they look like pictures. In practice, every image here was calibrated, aligned and processed to remove a much brighter star, its speckles and part of its disk. In 2014, researchers also had to subtract one planet to recover another.
That does not make the result less real. It makes the evidence cumulative. A molecular barcode rejects a dust clump. Radial velocity rejects the wrong motion. Eleven years of positions reject a stationary background object. Orbital modelling rejects many paths while showing where assumptions still matter. Atmosphere and evolution models translate light into a provisional physical world.
Beta Pictoris d did not suddenly enter the archive in 2025. A new observation supplied the trajectory that let astronomers find it there.
The same planet appeared twice in five days. Its discovery took 11 years of light.