Artist's impression of a Dyson bubble — a swarm of solar collectors surrounding a star

Dyson Sphere Candidates: Seven Stars That Glowed Too Bright — and What Was Behind Them

by
About 6 min read

In 1960, the physicist Freeman Dyson published a short paper in the journal Science that has haunted astronomy ever since. Dyson was not proposing a structure. He was proposing a search. If an advanced civilisation wanted to capture a large fraction of its star’s energy, he argued, it would have to build something around the star to catch the light — a shell, or more plausibly a swarm of collectors — and whatever it caught would have to go somewhere. The laws of thermodynamics are not negotiable: energy used becomes heat, and heat must be radiated away. A star surrounded by a civilisation’s machinery would therefore glow in the infrared, giving off more heat at long wavelengths than the star alone could explain. Look for stars that are too bright in the infrared, Dyson said, and you may find the engineers.

For sixty years, no one found one. Then, in May 2024, a team led by Matías Suazo and Erik Zackrisson at Uppsala University published a survey that came closer than anyone ever had. Project Hephaistos — named for the Greek god of the forge — had sifted through data on roughly five million stars and found seven that glowed with exactly the kind of infrared excess a partial Dyson sphere would produce. The internet heard “seven Dyson spheres”. The paper said something much more careful: seven candidates deserving further analysis. Two years later, the follow-up work was done, and the answer was not what the headlines had promised. It was, in its own way, just as interesting.

How to hunt a megastructure

The logic of the search is simple, and it is the reason Dyson’s idea has survived for more than six decades. A star’s light peaks at wavelengths determined by its temperature. A cool red dwarf, the kind of star the Hephaistos team was studying, puts out most of its light in the visible and near-infrared. If something around the star is intercepting a fraction of that light and re-radiating it as heat, the system will show an excess of mid-infrared light — light at wavelengths of ten microns or more — that the star itself cannot explain. That excess is the technosignature: the waste heat of a civilisation, visible across hundreds of light-years.

Project Hephaistos was built to find it at scale. The team combined data from three great sky surveys: Gaia DR3, the European Space Agency’s catalogue of more than a billion stars, with precise positions and distances; 2MASS, the Two Micron All-Sky Survey, which measured stars in the near-infrared; and WISE, NASA’s Wide-field Infrared Survey Explorer, which measured them in the mid-infrared, where waste heat would show. They ran about five million objects within 300 parsecs — roughly 980 light-years — through a filtering pipeline, using a convolutional neural network trained to spot the blending and confusion that plagues WISE’s low-resolution images. After cuts for nebulae, variable stars, bad astrometry and extended sources, and after visual inspection, seven objects survived every filter.

All seven were M dwarfs — small, cool red stars, the commonest kind of star in the galaxy — at distances of about 143 to 275 parsecs, or 470 to 900 light-years. In each case, the infrared light fitted a partial Dyson sphere at a temperature of 100 to 275 kelvin, capturing between 3 and 16 per cent of the star’s light. And in each case, the authors wrote, ordinary astrophysics struggled to explain the excess. Warm debris disks, very young stars, and background galaxies blended into the low-resolution WISE images were all possible explanations, and spectroscopy would be needed to tell them apart. The team was explicit: it was premature to call the excess a Dyson sphere. These were candidates, not detections.

The follow-up

The astronomical community did what it does with a good candidate list: it went and looked. And the looking, over the next two years, was thorough — and, for the megastructure hypothesis, unkind.

The first challenge came from the radio. In 2024, Tongtian Ren, Michael Garrett and Andrew Siemion cross-matched the seven candidates against all-sky radio surveys and found radio sources sitting 0.4 to 5 arcseconds from three of them — candidates A, B and G. Stars like these M dwarfs should not produce radio emission. A radio source next to a candidate was a strong hint that something else was in the line of sight: a background galaxy, shining through from far behind the star, whose light was being blended with the star’s in WISE’s blurry images.

High-resolution radio imaging settled one case decisively. Observations with e-MERLIN and the European VLBI Network, published in early 2025, resolved the radio source near candidate G into three compact components with a brightness temperature above a hundred million kelvin — the signature of a radio-loud active galactic nucleus, a supermassive black hole in a distant galaxy, sitting almost exactly behind the star from our point of view. And the imaging found no radio emission at the position of the M dwarf itself. Candidate G was not a Dyson sphere. It was a star with a galaxy hiding behind it.

A larger archival study, accepted by the Monthly Notices of the Royal Astronomical Society in 2026, compared Gaia star positions with WISE infrared centroids and deep images for the candidates. It found significant background contamination for candidates B and C, suggestive evidence for candidate A, marginal evidence for four others, and no obvious contamination for two — candidates D and I — which remained worth follow-up. The picture was becoming clear: some, and possibly all, of the seven candidates were contaminated by background galaxies rather than real stellar megastructures.

The James Webb Space Telescope finished the job for two of them. In 2026, JWST observations of candidates D and E — the two that the archival study had left standing — showed that their mid-infrared excess was likewise dominated by background contamination: dusty galaxies almost exactly aligned with the stars, not by megastructures, and not even by debris disks. The excess light was real. It was simply coming from somewhere else.

What the search actually found

As of late 2026, no Dyson sphere has been confirmed, and the best modern candidates appear to be background galaxies misclassified as stars. That sounds like a failure, and in the narrow sense — did we find aliens? — it is. But in the sense that matters to science, Project Hephaistos was a success, and a considerable one.

The first Hephaistos paper, published in 2022, set the strongest limits ever measured on how common partial Dyson spheres could be. Searching for structures absorbing between 1 and 90 per cent of a star’s light, with waste heat between 100 and 1,000 kelvin, the team found that among roughly 270,000 stars within 100 parsecs, the fraction hosting such structures was below about two in a hundred thousand. Among some 290 million stars within five kiloparsecs, the limit was below eight in ten thousand. Those are real numbers, and they mean something: if Dyson spheres exist in our neighbourhood, they are rare.

The second lesson is about method. The Hephaistos pipeline — five million objects, a neural network to reject confusion, careful cuts and visual inspection — is now the template for technosignature searches, and the follow-up campaign showed how quickly a candidate list can be tested: radio surveys, high-resolution interferometry, archival imaging, and JWST spectroscopy, all within two years. The search for alien engineering has grown up. It is no longer a thought experiment on the margins of astronomy. It is a mature observational programme, with instruments, pipelines, error bars and — crucially — a willingness to publish negative results.

And there is a third lesson, quieter than the others. The candidates failed because the universe is full of galaxies, and galaxies are bright in the infrared, and the sky is crowded. Every search for something extraordinary has to survive the ordinary first. The seven stars glowed too bright, and the reason was not a civilisation. It was the background, doing what the background always does: getting in the way, and reminding us that the most common explanation for a strange signal is a familiar object, seen badly.

What is documented, and what is not

Documented facts: In 1960, Freeman Dyson proposed in Science that an advanced civilisation capturing its star’s energy would re-radiate waste heat as infrared radiation, and that stars with unexplained infrared excess were worth searching for. Project Hephaistos I (Suazo, Zackrisson and colleagues, Monthly Notices of the Royal Astronomical Society 512:2988, 2022) searched about 270,000 stars within 100 parsecs and set an upper limit of below about two in a hundred thousand on the fraction hosting partial Dyson spheres. Project Hephaistos II (Suazo and colleagues, MNRAS 531:695, 2024) ran about five million objects from Gaia DR3, 2MASS and WISE through a pipeline including a convolutional neural network, and produced seven candidates, all M dwarfs at 143 to 275 parsecs, with infrared excess fitting partial Dyson spheres at 100 to 275 kelvin capturing 3 to 16 per cent of the star’s light. The authors described them as candidates requiring investigation, not detections. Follow-up found radio sources near candidates A, B and G (Ren, Garrett and Siemion, 2024); high-resolution e-MERLIN and European VLBI Network imaging showed the source near candidate G is a radio-loud active galactic nucleus, with no radio emission at the star (Ren and colleagues, 2025); an archival study accepted by MNRAS in 2026 found significant background contamination for candidates B and C; and James Webb Space Telescope observations in 2026 showed the excess of candidates D and E is dominated by background galaxies. As of late 2026, no Dyson sphere has been confirmed.

Unverified or disputed claims: That any of the seven Hephaistos candidates is a Dyson sphere or other alien megastructure — the authors never claimed this, and follow-up observations have traced the excess of several candidates to background galaxies. That Dyson spheres exist at all — no confirmed example is known, and the Hephaistos limits show they are rare in our neighbourhood if they exist. Alternative natural explanations, including warm debris disks and very young stars, remain possible for any candidate not yet resolved by follow-up.

Sources

Leave a Comment