Shutting down a nuclear reactor isn’t much like flipping off a light switch. The control rods go in and the chain reaction stops within seconds, but the fuel is still packed with radioactive leftovers from all that splitting, and they keep decaying for months, sometimes years. That’s why used fuel spends several years at the bottom of a cooling pool before anyone ships it off.<\/p>\n
That decay usually comes up as a heat problem. But it turns out those leftovers are also throwing out antineutrinos, and a team of physicists has now published a count of them, taken at the Chooz nuclear plant in France’s Ardennes region with both of its reactors switched off. According to the Double Chooz collaboration, it’s the first time anyone’s put a real number on that signal.<\/p>\n
If you slept through that part of high school physics, an antineutrino is the antimatter version of the neutrino. It’s so light and so antisocial that it’ll pass through a reactor building, the ground under it and you while almost never bumping into anything. Reactors make enormous numbers of them, because many of the radioactive fragments inside the fuel decay by spitting one out.<\/p>\n
That antisocial streak makes them a nightmare to catch, and it’s also why nuclear inspectors have been interested in them for a couple of decades now. Concrete does a fine job on the radiation you’d normally worry about, which is why there’s four feet of it around the argon-filled room in Idaho where US researchers cut open used fuel rods<\/a>. An antineutrino goes through that wall like it isn’t there. You can’t shield them, and you can’t fake them.<\/p>\n You build a big, well-shielded tank of liquid, and then you wait. The Double Chooz near detector sits underground about a quarter mile from the plant’s two reactor cores, and its innermost tank holds 2,721 gallons of liquid scintillator spiked with gadolinium. Scintillator’s basically a liquid that gives off a tiny flash of light when a particle dumps energy into it.<\/p>\n Another 5,970 gallons of scintillator without the gadolinium surrounds that tank, followed by a 41-inch layer of mineral oil with 390 ten-inch photomultiplier tubes pointed inward. A photomultiplier tube works a bit like a light bulb in reverse: a speck of light goes in and an electrical signal comes out. And the near detector had about three feet of water around it to soak up radiation coming off the surrounding rock.<\/p>\n When an antineutrino does smack into a proton in the liquid, the collision produces a positron and a neutron. The positron makes a flash right away. The neutron bounces around for a split second before something grabs it, often a gadolinium atom, and that capture makes a second flash. Thierry Lasserre, one of the two Max Planck Institute for Nuclear Physics researchers who led the study, described it in the institute’s release as “a characteristic double-light signal.” Double Chooz only counted pairs where the second flash showed up within 800 millionths of a second and about four feet of the first. That’s frankly a pretty clever way to sort real hits from random junk, since radioactivity in the tank walls doesn’t usually flash twice on cue.<\/p>\n You might be wondering why this took so long. Chooz B has two reactors, and the plant staggers their refueling, so it’s pretty rare for both to be down at once. All of the data comes from 2017, when both cores were shut down together on four separate occasions for refueling or maintenance.<\/p>\n Those four windows added up to 24.4 days. Once you subtract the dead time the detector racks up whenever a cosmic-ray muon (a heavier cousin of the electron that rains down out of the sky) sets off its veto, you’re left with 17.2 days of usable data from the near detector.<\/p>\n So what did it see?<\/p>\n In the low-energy range where the leftover signal should show up, the near detector logged 244 candidate events. More than half of those were background, and once that’s subtracted you get 106 events, give or take 18. Simulations that tracked the history of every fuel assembly in the cores and the pools had predicted 88, give or take 7, and that’s a pretty tidy match.<\/p>\n Physicists usually want 5 sigma before they’ll call a signal real, and this one comes in at 5.9. For comparison, the same near detector recorded about 900 antineutrino events a day with the reactors at full power. With both off, it averaged a little over six a day from the leftovers.<\/p>\n The far detector, about two-thirds of a mile from the cores, picked up 27 extra events where the model expected 14. That’s in line with the model too, just with a lot more wobble, since the signal drops off quickly with distance.<\/p>\n Arguably the more interesting number is where that signal came from. The collaboration’s model puts about 56% of it on fuel still sitting inside the two shut-down cores and 44% on used fuel in the cooling pools, which are in a building roughly 125 feet from the reactors. They’re the same kind of pools fuel sits in between jobs, like the bundles that went from a pool back into a reactor on the Lake Michigan shore<\/a>.<\/p>\n Fuel in a freshly shut-down core fades fast, because a lot of what’s decaying in there dies off within hours. The pools don’t work that way. In the energy range Double Chooz measured, most of the signal comes from praseodymium-144 and rhodium-106, which only last minutes and seconds. But they’re constantly topped up by longer-lived parents: cerium-144, with a 285-day half-life, and ruthenium-106, at 372 days. Which means fuel that’s been cooling for a year or two keeps putting out antineutrinos long after it last saw the inside of a reactor.<\/p>\n Mostly nuclear safeguards. The International Atomic Energy Agency is the organization that checks whether uranium and plutonium at power plants stay where they’re supposed to, and for stored spent fuel, that checking’s still largely a matter of accounting, according to the paper. A detector that can read the fuel through the walls, even with the reactor off, would give inspectors a check that doesn’t depend on anyone’s paperwork. The Double Chooz team calls its result a proof of principle for exactly that.<\/p>\n The American Physical Society’s Physics magazine<\/a> ran its summary of the study under the headline “Detecting the Illicit Removal of Nuclear Fuel,” and that’s a bit ahead of what the paper claims. The collaboration writes that its measurement can confirm the leftover signal’s there and would notice if the total amount of stored fuel changed by a lot, but it isn’t sensitive enough to spot a small number of missing fuel assemblies. For scale, each Chooz B core holds 205 assemblies, and each one carries roughly 1,300 pounds of uranium dioxide.<\/p>\n The detector isn’t anything you’d carry around, either. The authors say this result lets people set realistic goals for smaller, cheaper, more portable detectors, and they point to spent fuel pools and dry cask pads (the concrete slabs where older fuel ends up sealed in big steel and concrete casks) as the most natural places to use them. They admit the hardware would still be sizable, but they don’t put a size or a price on it. I’d guess no one’s rolling one of these up to a storage pad any time soon, and the paper says keeping tabs on a deep underground repository this way is a long way off.<\/p>\n The underlying data isn’t public, so I can’t rerun any of this myself, though the authors say they’ll share it on reasonable request. Other teams are already chasing the same signal, too: Max Planck’s release on the study<\/a> says JUNO-TAO, a detector about 144 feet from a reactor at China’s Taishan plant, is using reactor-off data to try to isolate the spent fuel signal, and that its first results were shown at the Neutrino 2026 conference. The release doesn’t include TAO’s numbers.<\/p>\nSo how do you catch something that goes through everything?<\/h2>\n
Both reactors had to be off at the same time<\/h2>\n
Almost half of it came from the cooling pools<\/h2>\n
So what’s this actually good for?<\/h2>\n