OrbitalHub

A small team in Uppsala, Sweden spent five years winnowing a million stars down to nine that glowed strangely warm in the mid-infrared. The signal looked exactly like the signature Freeman Dyson predicted in 1960: starlight absorbed by a swarm of engineered habitats, the waste heat re-radiated at longer wavelengths. When the James Webb Space Telescope finally turned its 6.5-meter mirror on the two most promising candidates this summer, the answer came back in redder light than anyone expected. Both stars were ordinary M dwarfs. The infrared excess came from galaxies nearly a billion light-years behind them. Project Hephaistos had bagged two Hot DOGs and a dusty starburst, not a Kardashev-II civilization.

The search for extraterrestrial intelligence has spent most of its history chasing radio. That made sense in 1960, when Cocconi and Morrison published “Searching for Interstellar Communications” and Frank Drake pointed the Green Bank telescope at Tau Ceti. Radio is cheap, narrow-band, and physics says it can carry information across the galaxy. Optical SETI followed, then gamma-ray and neutrino lookups, then the more recent push for “technosignatures,” a deliberately broad term for any measurable fingerprint of engineered technology.

Dyson signatures sit at the far end of that spectrum. A civilization that wants to harvest a non-trivial fraction of a star’s luminosity has to build something physically large: a swarm of collectors, a shell, a statite array. No known natural astrophysics produces a star that is unusually faint at visible wavelengths while flooding the mid-infrared. The contrast is the whole point.

That contrast also makes Dyson searches almost uniquely easy to spoof. Dust around young stars looks like Dysons. Active galactic nuclei look like Dysons. Gravitationally lensed background galaxies look like Dysons. The trick is not finding candidates. The trick is surviving the long chain of false positives.

Project Hephaistos started as a Stockholm-Uppsala collaboration in 2021, led by Erik Zackrisson. The first step was cross-matching the Gaia DR3 catalog with WISE infrared photometry, then applying color cuts that ought to leave only stars whose mid-infrared output dwarfs their visible output. A first paper in 2024 narrowed roughly one million M dwarfs within 300 parsecs down to seven that survived. A second paper added two more. Nine candidates, designated A through I, each a small red star, each a few billion years old, each suspiciously dim in the optical.

The follow-up paper, Hephaistos III in July 2026 (Korn et al., arXiv:2607.25701), spent months chasing the candidates with ground-based photometry and low-resolution spectroscopy. None of them showed the H-alpha emission or lithium absorption lines expected of a young star with a transitional circumstellar disk. Stellar activity could explain some of the variability, but not the bulk of the infrared excess. The team couldn’t kill the candidates, but they couldn’t confirm them either.

Then Webb settled it.

Hephaistos IV (Zackrisson et al., submitted to MNRAS, arXiv:2607.09460) targeted two of the strongest candidates, D and E, with MIRI imaging and spectroscopy in June 2026. Candidate D, the most exciting, sat at a redshift of roughly z = 0.9, appeared point-source-like in the imaging, and showed a mid-infrared spectrum consistent with a Hot Dust-Obscured Galaxy. Hot DOGs are some of the most luminous galaxies in the universe: their central engines are buried under thick blankets of warm dust that reradiate the optical and ultraviolet output into the infrared, exactly the spectrum a Dyson swarm would produce if you couldn’t resolve it.

Candidate E was a z ~ 0.4 galaxy with extended morphology, bright knots of star formation, and the spectral signature of a dusty starburst. Both galaxies sit within roughly an arcsecond of their foreground M dwarfs on the sky. From WISE’s coarse pixels, the starlight and the galaxy light smeared together. To WISE, that smear looked like a faint optical star with too much infrared emission. To JWST, it looked like an M dwarf with a galaxy photobombing it.

The Hephaistos team’s conclusion is dry: the infrared excess does not originate from Dysonian megastructures or any other radiation mechanisms close to these stars. The next round of search will need much higher spatial resolution than WISE can offer. The 2030s infrared observatories that Andreas Korn and his colleagues point to in their press materials are designed to do exactly that.

The Dyson concept sits inside a framework the Soviet astronomer Nikolai Kardashev proposed in 1964. A Type-I civilization uses all the energy available on its home planet, roughly 10^16 watts for Earth. Type II harvests its star, on the order of 10^26 watts for the Sun. Type III commands an entire galaxy, around 10^36 watts. Dyson’s original 1960 paper in Science asked how a Type-II civilization would dispose of waste heat, and noted that any structure intercepting a meaningful fraction of a star’s luminosity has to reradiate that energy somewhere. The mid-infrared is the natural choice, because a swarm’s equilibrium temperature, set by balancing absorbed starlight against emitted blackbody radiation, falls somewhere around 200 to 400 kelvin for a star like the Sun.

The peak wavelength of that emission is set by Wien’s displacement law, λ_peak (μm) ≈ 2898 / T (K).

A 300 kelvin Dyson swarm peaks near 9.7 micrometers, right inside WISE’s W3 band at 12 micrometers and JWST MIRI’s red end. That’s why the Hephaistos search uses W3-W1 color cuts: stars that should look like blackbodies at 2,800 to 3,800 kelvin but instead dump energy into a 9-micrometer band stand out.

The contamination problem is structural. WISE’s 12-micrometer pixels are 6.5 arcseconds wide. JWST’s shortest MIRI imaging band runs near 5 micrometers with diffraction-limited resolution of about 0.1 arcseconds. The angular resolution ratio is roughly 65 to 1, which means WISE blends sources that JWST can split. A faint background galaxy with a hot dusty nucleus will add flux to every pixel that contains the M-dwarf centroid. If the galaxy is at high redshift and K-corrected into the mid-infrared, its spectrum can look almost like a blackbody with the right temperature. That’s exactly what Hot DOGs do.

For the candidate identification, the math is straightforward. Given an apparent magnitude in the visible m_v and a magnitude in the mid-infrared m_W3, a star with a Dyson shell at temperature T emitting as a modified blackbody shows an infrared excess of Δ(m_W3 − m_v) = (m_v − M_v) − (m_W3 − M_W3) − BC_corr, where M_v and M_W3 are the absolute magnitudes expected for the stellar photosphere alone. Anything more than a few magnitudes of excess on top of the expected Rayleigh-Jeans tail triggers the flag. WISE picked up roughly 9 such objects out of 10^6.

The engineering is also telling. JWST’s MIRI instrument, which did the actual follow-up, is a 50-kelvin cryocooled camera and spectrograph operating from 5 to 28 micrometers. It took roughly 30 years from concept to first light. The contrast with WISE is sharp: WISE was a 40-centimeter survey telescope that mapped the entire sky in six months, while JWST is a 6.5-meter segmented mirror at L2 that can spend hours on a single target. The two designs are complementary, and that complementarity is exactly what broke this case open.

The Dyson search is not getting easier. It is getting slower, more expensive, and more rigorous. Each round of follow-up costs more telescope time than the round before, and the surviving candidates shrink in number faster than new surveys add to the list. The Hephaistos result is part of a pattern: most candidate Dysons turn out to be dusty galaxies, AGN, or stellar variability. A few percent have stayed unexplained, but the unexplained share keeps shrinking.

For the Dyson concept itself, none of this is fatal. A Type-II civilization would not necessarily produce a signature WISE could see, and a Type-III civilization’s footprint is so diffuse that it would look like background noise in any single survey. The Hephaistos team is already planning the next iteration, with higher-resolution mid-infrared facilities in the 2030s and follow-up on the surviving seven candidates that Webb hasn’t yet looked at. The question that motivated Freeman Dyson in 1960, where the waste heat of an advanced civilization ought to show up, still has no answer. We are simply narrowing where it isn’t.