In 1972 France received uranium that had already run

In June 1972, at the Pierrelatte plant in France's Rhône valley, a technician in the control laboratory measured a flask of uranium hexafluoride —the gas from which nuclear fuel is prepared— and got 0.717% uranium-235 where it should have read 0.720%. Three thousandths of a point: thirty atoms in every million. Three months later France was accepting the only explanation compatible with that shortfall: the uranium had come from a deposit in Gabon that had run as a nuclear reactor on its own, 1.7 billion years ago and for hundreds of thousands of years.

That 0.720% is one of the most stubborn figures there is. Uranium-235, the isotope that splits and releases energy, sits at exactly that proportion in the granite of Galicia, in the sand of the Sahara and in the rocks Apollo 11 brought back from the Moon. That is why the measurement works as a standard: when a flask deviates, the instrument is usually at fault.

The instrument was fine. They repeated the measurement on a second spectrometer and got the same answer. They traced the batch backwards, drum by drum, to the Oklo deposit in south-eastern Gabon, inside the same mining concession as Mounana, the mine that closed in 1999, worked out. There were veins there with 0.44% uranium-235, and in places even less. Fissile uranium was missing, and a great deal of it.

In September it was presented by the physicist Francis Perrin to the French Academy of Sciences, and what his report rested on was not the uranium that had disappeared, which allows chemical alibis, but what sat in its place. Oklo turned up neodymium and ruthenium with a blend of isotopes found in no mineral on Earth, and found exactly, on the other hand, in the residue of spent fuel. Those two elements are the ash of fission, and there they were.

The reason is the arithmetic of half-lives. Uranium-235 decays with a half-life of 704 million years; uranium-238, its majority sibling, takes 4,470 million. The fissile one burns down six times faster. One point seven billion years ago, the natural uranium of the whole planet was not at 0.7% but close to 3%, roughly the enrichment a commercial nuclear plant uses today. All that was missing was the water.

Water is the part that turns a seam into a machine. A neutron fresh out of a split atom travels too fast to split the next one: it has to be slowed down, and the hydrogen in water does that well. At Oklo, groundwater soaked the sandstone, moderated the neutrons, and the reaction started on its own. Then the heat boiled the water off, the moderator left as steam, and the reaction stopped. The rock cooled, the water came back in, and it all began again.

In 2004 Alex Meshik and his team at Washington University in St Louis measured the xenon trapped in the grains of that rock and worked out the cycle: about 30 minutes on and about two and a half hours off, for hundreds of thousands of years. The average power came out at around 100 kilowatts — what fifty electric radiators draw at once.

Sixteen reaction zones have been found in the Franceville basin: at Oklo, at Okélobondo and at Bangombé. And one detail made them useful. The Bangombé zone lies twelve metres down, almost exposed, and its fission products —plutonium included, which decayed on the spot— are still inside the rock that made them. They did not move in 1.7 billion years. That is why the International Atomic Energy Agency studies Oklo as the only long trial that exists of a nuclear waste repository: the experiment has already been run, it lasted a thousand times longer than any budget, and the samples are numbered.