For 17 days both reactors at a French plant sat completely cold, and a buried tank of clear liquid a quarter mile away still counted about a hundred ghost particles coming off the spent fuel

Both units at the plant were down.

No fission, no steam, nothing on the wires, nothing in either core doing what a core does.

A quarter mile away and well under the ground, a tank of clear liquid kept counting.

It was built to pin down theta 13, a mixing angle in neutrino physics, by comparing antineutrinos from the two reactors at Chooz in northern France at two distances.

Not many. About six a day.

But six a day out of a machine that had gone quiet is a number, and a number is the entire point.

The fuel had stopped making energy. It had not stopped telling on itself.

Why a dead core is not a silent one

Fission stops within seconds of a shutdown. What fission left behind does not.

The fuel is loaded with fission products, hundreds of unstable nuclides, and they keep decaying on their own schedules for days, years and centuries.

Most of those are beta decays, and every beta decay throws out an electron antineutrino.

That particle is why any of this works. It interacts with matter so weakly that steel, concrete and miles of rock are effectively not there.

Which cuts both ways. Nothing blocks the signal, and nothing catches it either.

What a detector can do is offer it enough protons to run into. Once in a very long while an antineutrino hits one and converts it, producing a positron and a neutron.

The two arrive microseconds apart and leave two flashes in sequence, and that double signature separates a real event from everything else making light down there.

The machine, and what it was actually built for

The detector is called Double Chooz, and shutdown fuel was never its job.

It was built to pin down theta 13, a mixing angle in neutrino physics, by comparing antineutrinos from the two reactors at Chooz in northern France at two distances.

The near tank sits a quarter mile from the cores, under enough rock to equal roughly 375 feet of water, which is what it takes to hold off the cosmic ray muons.

Inside is about 360 cubic feet of liquid scintillator carrying a gram of gadolinium per liter.

The gadolinium is there to grab the neutron and make that second flash unmistakable.

Each core runs at 4.25 gigawatts of thermal power, which is a great many antineutrinos to work with, and nothing on the instrument had to change to go looking for the quiet ones.

What seventeen days of nothing produced

The collaboration had 17.2 days with both units genuinely off.

In that window it counted 106 candidate events, give or take 18.

A simulation of what should have been left inside the fuel predicted 88, give or take 7.

Those two overlap, and the whole result rests on that overlap. It puts the detection at 5.9 sigma, well past the threshold physics is willing to call a discovery.

The signal concentrates between 1 and 3 MeV, and two decay chains dominate there. Praseodymium 144 accounts for something over half of it, rhodium 106 for close to another 40 percent.

The source is not only the cores. A large part comes from the spent fuel sitting in the cooling ponds beside them, which is why the paper is written up as a measurement of spent fuel rather than of a shut reactor.

The part that does not scale

A result is not an instrument you can ship.

Double Chooz works because it is buried under a particular hill next to a particular plant, and because the team spent years characterizing everything else down there that makes light.

Cosmic ray muons, neutrons correlated with the reactors themselves, natural radioactivity in the surrounding rock.

All of it had to be pinned down before six events a day could be pulled out cleanly, and at a different plant that campaign starts again from close to zero.

Then the shielding. Putting the equivalent of this beside every facility under international inspection is not a budget anybody has.

The published measurement is explicit that it is a demonstration. Work on water based and doped scintillators is aimed at something smaller, but a container sized detector with this sensitivity does not exist yet.

What it would mean for an inspector

Safeguards rest on operator declarations, seals, cameras and periodic measurements of fuel.

Those work when a plant is running and a core is accessible.

A reactor shut for months, with its fuel in a pond nobody is opening, is harder to read from outside the fence.

The paper points at two uses in particular. Verifying the spent fuel inventory, and estimating residual power.

Both fall out of the same relationship. More burnup means more long lived fission products, which means a stronger and longer tail, and that relationship runs in one direction only.

The accounting problem underneath is real enough that undeclared uranium keeps surfacing in cargo nobody declared it in, while spent fuel management has its own policy fight alongside.

None of that is settled by 106 events in a tank in the Ardennes. What those events settle is narrower. A reactor with nothing happening inside it is still readable, and nobody had shown that before.