Power down a nuclear reactor and the core doesn’t fall silent. It merely fades.
Inside the fuel, long-lived radioactive fission products go on decaying for months and sometimes years, releasing a thin stream of antineutrinos that slips straight through the pressure vessel, the shielding and anything else standing in the way. That leftover trickle had never been measured by anyone. The Double Chooz collaboration has now done it, drawing on 17.2 days of data taken while both reactors at the Chooz plant in northern France were completely shut down.
Across those 17.2 days, the tally came to roughly 100 antineutrino candidate events — around six per day — picked up by a detector filled with more than 30 cubic meters of liquid scintillator, sitting underground roughly 400 meters from the two reactor cores.
Why almost nobody bothers with reactor-off data
Of all the particles we know about, antineutrinos are the lightest and the hardest to pin down, streaming through a reactor and its shielding with next to no interaction. That property is simultaneously what makes them valuable and what makes them so wretched to detect.
Reactors that are running spew out a torrent of them, and that is why decades of reactor antineutrino physics have been aimed at cores in operation. Set against that torrent, the shutdown signal is a rounding error.
“Antineutrinos interact only extremely rarely with matter. However, when one interacts within the Double-Chooz detector, a characteristic double-light signal is produced that can be distinguished from background events,” said Thierry Lasserre of the independent research group OMINA, based at the Max Planck Institute for Nuclear Physics in Heidelberg, Germany. Everything hinges on that double flash. Strip it away and 100 events spread over two and a half weeks simply disappear into the noise.
The simulation got it right, which is the actual news
What the detector saw lined up with detailed simulations that folded in the remaining nuclear fuel inventory along with the decay of long-lived fission products. It is the first direct experimental confirmation of predictions that had existed only on paper until now.
That result carries more weight than the detection itself. A predictable signal can serve as a baseline, and a baseline is the thing that converts a physics result into an instrument.
The events originate in residual radioactivity, both in the cores themselves and in the spent-fuel cooling pools nearby. Each source contributes. Each was modeled.
Years of background work, not a lucky run
“Until now, reactor antineutrino experiments have mainly focused on operating reactors, where the antineutrino flux is much larger. Detecting the tiny residual signal after shutdown required exceptionally low backgrounds and careful analysis techniques developed by the Double Chooz collaboration over many years,” said Anthony Onillon, who led the work alongside Lasserre at the Max Planck Institute for Nuclear Physics. The paper appeared recently in Physical Review Letters.
Double Chooz no longer has this territory to itself. Initial results from JUNO-TAO, presented at Neutrino 2026, show researchers turning to reactor-off data to examine the faint antineutrino signal coming from spent nuclear fuel, though TAO is still working to isolate that emission. Double Chooz landed the first published benchmark for the residual signal from shut-down reactors and spent-fuel pools.
What a safeguards inspector might do with this
Reactor monitoring is the practical selling point. A detector able to register a core during maintenance and after shutdown — not solely at full power — offers a route to independently verifying reactor status and following spent-fuel inventories without relying on anyone’s word.
The operative word is could. This amounts to 100 events from a single detector at a single plant, 400 meters out from the cores, and what the paper establishes is that the signal is there and behaves the way it was predicted to. Converting that into a deployable safeguards tool is a separate engineering challenge, and the research makes no claim otherwise.
The trajectory is clear enough all the same. Shielding, paperwork and declarations mean nothing to antineutrinos, and that is precisely the property any verification regime is looking for.
The detector was built to do something else entirely
The original purpose of Double Chooz was investigating neutrino oscillations, and it played a key role in measuring the neutrino mixing angle θ13 — the parameter that describes how neutrinos shift from one type to another in flight. That measurement helped open the door to research into matter-antimatter asymmetries in the neutrino sector.
The same underground tank has now logged a second first: the glow that keeps seeping out of a reactor once the reactor stops.
If one number is worth remembering, make it 17.2 days. That is how long two idle reactor cores had to sit there, doing nothing anyone could perceive, before physics 400 meters away could demonstrate that they weren’t finished.