Seventeen days of data taken while both reactors at Chooz were off yielded about a hundred antineutrino events from spent fuel and decaying fission products.
A nuclear reactor does not stop emitting neutrinos when it is switched off. Fission products built up during operation keep decaying for years, and every beta decay releases an electron antineutrino. The core keeps glowing in a radiation nothing can shield, and so does the pool of spent fuel beside it.
That residual emission has been predicted for decades and never measured. A paper in Physical Review Letters from the Double Chooz collaboration reports the first measurement of it, using seventeen days of data taken while both reactors at the Chooz plant in northern France were fully shut down.
The work was co-led by Anthony Onillon and Thierry Lasserre of the Max Planck Institute for Nuclear Physics in Heidelberg, the latter heading the institute’s independent OMINA research group.
Measuring what is left over
The Double Chooz detector sits underground about 400 metres from the plant’s two reactor cores and contains more than 30 cubic metres of liquid scintillator. It was built for a different purpose — studying neutrino oscillations, and it played a central role in measuring the mixing angle θ13 — which is why a detector capable of this measurement happened to be pointed at a reactor that would eventually be turned off.
The analysis covers 17.2 days during which both units were down. In that window the detector recorded about 100 antineutrino candidate events, originating in residual radioactivity in the reactor cores and in the nearby spent-fuel cooling pools.
“Until now, reactor antineutrino experiments have mainly focused on operating reactors,” Onillon said.
The signal agrees well with detailed simulations of the remaining fuel inventory and the decay of long-lived fission products, which the collaboration describes as the first direct experimental validation of those predictions.
Why a shutdown reactor is hard to see
The difficulty is not conceptual but statistical, and it comes down to how feebly these particles interact.
“Antineutrinos interact only extremely rarely with matter,” Lasserre said.
An operating gigawatt reactor emits something on the order of 1020 antineutrinos a second, and a detector 400 metres away with tens of cubic metres of target catches a few per hour. The residual emission from a shut-down core is smaller by orders of magnitude, which is why the measurement needed a detector already built for precision oscillation work, sitting underground to suppress cosmic-ray background, and why a hundred events over seventeen days counts as a result.
A hundred events is also, plainly, a hundred events. The statistical uncertainty on a sample that size is around ten per cent before any systematic uncertainties are considered, so this is a detection and a consistency check rather than a precision measurement of the residual spectrum.
The reason anyone wants this number
The practical interest is in safeguards. Antineutrinos pass through everything, cannot be screened or spoofed, and carry information about what is inside a reactor — which has made neutrino monitoring a long-standing proposal for verifying that reactors and fuel stores contain what their operators say they contain.
Spent fuel is the part of that picture where the proposal has been weakest, because the predicted signal had never been checked against a measurement. Spent fuel is also where the proliferation concern actually lives: plutonium is in the fuel after irradiation, not in it beforehand, and a cooling pool holds material that a safeguards regime would like to be able to count without opening anything.
A validated model of the residual emission matters for operating reactors too. Any precision measurement at a working reactor has to subtract the contribution from decaying fission products and from the pool next door, and until now that subtraction rested on calculation alone. Double Chooz provides the first published benchmark for it.
What the measurement does not settle
Several limits are worth stating, and the release states none of them.
The agreement is with a simulation of this particular plant’s fuel inventory — its burn-up history, its cooling times, the contents of its pools. That the model works here is good evidence that the underlying nuclear data are sound, and it is not a demonstration that an inventory could be inferred from the signal without knowing the answer in advance. Validating a prediction and performing an inverse measurement are different problems, and safeguards needs the second.
The event count also does not separate the core from the pools. Both sit within a few hundred metres of the detector, and the release reports a combined signal consistent with a combined prediction, not two independently resolved sources.
And a shutdown long enough to collect clean data is a rare opportunity rather than a method. Reactors are not switched off for the convenience of particle physicists, which is part of why this measurement waited so long.
Others are now looking
Double Chooz is no longer alone in this. First results from JUNO-TAO, presented at the Neutrino 2026 conference, show other experiments pursuing the same measurement — JUNO-TAO being a high-resolution detector placed close to a commercial reactor in China specifically to characterise the reactor antineutrino spectrum.
That convergence is the useful signal here. A single hundred-event detection from a detector built for something else is a first; two or more independent measurements of residual emission, with better statistics and better separation of sources, is the beginning of a technique.
The release does not state a DOI for the paper, and does not state the study’s funding sources or the full author list. The page also spells Lasserre’s surname two different ways.
Notes
- “First measurement of neutrino emissions from spent nuclear fuel by the Double Chooz experiment.” Physical Review Letters, 2026. The release does not state a DOI.
- Max-Planck-Institut für Kernphysik. “First measurement of antineutrinos from spent nuclear fuel.” EurekAlert!, 4 August 2026. https://www.eurekalert.org/news-releases/1138527.
Photo by Markus Distelrath on Pexels.com