In May 1972, at France's Pierrelatte nuclear fuel processing plant, a routine check made the duty technician frown.
Ore samples shipped from the Oklo uranium mine in Gabon showed a uranium-235 abundance of 0.7171% — not 0.7202%, the standard value for all natural uranium ore. A difference of 0.003 percentage points.
In the nuclear industry, such a deviation means one of two things: either someone stole U-235 to build a bomb, or the measurement was wrong. The technician re-measured. Still 0.7171%.
The news went up to the French Atomic Energy Commission. Physicist Francis Perrin took over the investigation. He examined samples from different locations and depths in the Oklo mining district and found that the U-235 deficit was not uniform — some ore layers were missing absurd amounts, with the lowest at just 0.44%.
This was not theft, and it was not error. Perrin did something that stunned his colleagues: he compared the isotopic spectrum of Oklo ore with the spent fuel of modern nuclear power plants.
Almost identical.
The only explanation: this uranium had already been fissioned.
A Prediction Abandoned for 16 Years
The story begins in 1956. That year, nuclear chemist Paul Kuroda of the University of Arkansas published a paper with a mundane title — *On the possibility of neutron-induced chain reactions in uranium minerals*. He listed four conditions that a natural nuclear reactor must satisfy simultaneously:
First, the U-235 abundance must be high enough. U-235 has a half-life of about 700 million years; U-238, about 4.5 billion years. The further back in time, the higher the U-235 proportion. Kuroda calculated that about 2 billion years ago, natural uranium reached roughly 3% U-235 — exactly the enrichment level of modern pressurized water reactor fuel.
Second, the uranium deposit must be large and dense enough to form a "critical mass."
Third, there must be a moderator — a substance that slows neutrons. Neutrons released by fission travel too fast; hitting a uranium nucleus directly just bounces off, like throwing a ping-pong ball at a wall. Something must "brake" the neutrons to thermal speeds so uranium nuclei can capture them and sustain the chain reaction.
Fourth, the deposit must be surrounded by a sealing structure to prevent neutrons from escaping.
After Kuroda's paper appeared, almost nobody paid attention. A natural nuclear reactor? Sounded like science fiction. Sixteen years later, Oklo's ore answered for him.
Water: The Oldest Safety Valve
How did the Oklo reactors actually run? The question puzzled scientists for years — until 2004, when Alex Meshik's team at Washington University delivered a precise answer.
They analyzed not uranium, but xenon.
Xenon is one of uranium's fission products, with nine isotopes whose half-lives range from seconds to millions of years. Meshik found "the highest concentration of xenon known in nature" inside Oklo's aluminum phosphate mineral grains. This xenon had been locked in atomic-scale holes in phosphate crystals, preserved like insects in amber for 2 billion years.
Strangely, though, the xenon was not in the uranium minerals but in phosphate grains scattered through the reactor-zone rock. This meant the xenon was captured only while the reactor was cooling — only then could xenon diffuse out of the uranium minerals and be "frozen" into the phosphates.
By measuring the relative proportions of different xenon isotopes, Meshik reconstructed the reactor's operating rhythm: one cycle every 3 hours — 30 minutes of operation, 2.5 hours of cooling.
The director behind this rhythm was water.
Groundwater seeped along rock fractures into the uranium ore body, filling the gaps between ore grains. The hydrogen atoms in water are nearly the same mass as a neutron — a natural braking pad. A neutron striking hydrogen slows abruptly, from fast neutron to thermal neutron, just right for capture by U-235. The chain reaction starts, and temperature rises.
Then the water boils.
Steam escapes, the moderator disappears. Neutrons speed up again, are no longer captured, and the chain reaction stops. The reactor cools. Groundwater seeps back in. The moderator returns. The chain reaction restarts.
30 minutes of boiling, 2.5 hours of waiting for the water to return. This cycle continued for roughly 150,000 years.
This is the oldest automatic control system known. No sensors, no actuators, no PID controller. Just one physical law: water boils at 100°C. That simple, that reliable.
The Time Window: Why Only Oklo
Natural reactors were not uniquely possible at Oklo. The harshest of Kuroda's four conditions was the first — U-235 abundance.
Two billion years ago, U-235 made up about 3% of natural uranium. Today it is 0.72%. More than four times less. The reason is simple: U-235 decays over six times faster than U-238, halving every 700 million years. Two billion years ago sat exactly in a "golden window" — a few hundred million years earlier, U-235 was too abundant and reactors would risk runaway; a few hundred million years later, too scarce and the chain reaction could never ignite.
That window is permanently closed. No new natural reactors can form on Earth. U-235 abundance will only keep declining, until the Sun swallows the planet.
Oklo happened to stand at the very center of the window. Gabon's geology supplied the other three conditions: ore large and dense enough, groundwater as moderator, and surrounding sandstone and granite sealing in the neutrons. Scientists have identified 16 reactor zones at Oklo; the largest consumed about 6 tonnes of U-235, running at an average output of roughly 100 kW — about 1,000 light bulbs, or one electric car.
A 2-Billion-Year Nuclear Waste Experiment
Oklo's most valuable legacy is not the reactors themselves, but their waste.
Nuclear waste disposal is one of the industry's hardest problems. High-level waste has half-lives of tens of thousands of years; human civilization is only a few thousand years old. How do you guarantee a repository stays safe after 100,000 years? Geologists can model, engineers can design, but nobody can run the experiment. Humanity cannot wait 100,000 years.
But Oklo could.
After the reactors shut down, fission products were sealed by the surrounding sandstone, granite, and clay. Over 2 billion years, the most dangerous nuclide — plutonium — migrated less than 3 meters. Three meters. The distance from your desk to the door. In 2 billion years, continents drifted thousands of kilometers, mountains rose and eroded, glaciers came and went — and plutonium moved 3 meters.
This data directly shaped global nuclear repository design philosophy. If Oklo's rock could hold plutonium for 2 billion years, then a carefully chosen geological environment plus engineered barriers should be safe on a 100,000-year scale — at least far safer than previously assumed.
A Reactor That Answered a Cosmology Question
The story gets even wilder.
In 1976, Soviet physicist Alexander Shlyakhter pointed out that Oklo's samarium-149 neutron capture data could test a fundamental physical question: does the fine-structure constant α change over time?
The fine-structure constant α ≈ 1/137 is the dimensionless constant describing the strength of electromagnetic interaction. If α changed over cosmic history, the foundations of modern physics would shake — from quantum electrodynamics to grand unified theories, everything would need rewriting.
Samarium-149's neutron capture cross-section is exquisitely sensitive to α. During reactor operation, Sm-149 absorbed neutrons like a sponge, and its depletion recorded the cross-section at the time. By measuring Sm-149 to Sm-147 ratios in Oklo ore, one can infer the value of α 2 billion years ago.
Decades of analysis by multiple teams conclude: α has changed by no more than one part in ten million (0.1 ppm) over 2 billion years.
A natural nuclear reactor became a "time machine" for testing whether cosmic constants are constant. Kuroda never dreamed of this in 1956.
Aftertaste
The Oklo story has three layers, each deeper than the last.
The first is wonder: Earth built its own nuclear reactor, 1.7 billion years before humans.
The second is engineering wisdom: the water-heat feedback loop is the simplest automatic control system ever — one negative feedback loop, no parts, nothing to break, running for 150,000 years. Human reactor safety systems have redundancy, backups, defense-in-depth, but the underlying logic is the same as Oklo's: temperature rises → negative feedback engages → power drops. The difference is that Oklo's "sensor" and "actuator" are the same thing — water.
The third is humility about timescales. We worry about waste safety over 100,000 years; Oklo already ran the experiment for 2 billion. We ask whether cosmic constants are eternal; Oklo's samarium isotopes answered to within 0.1 ppm. Sometimes our most precise instruments cannot outdo a rock that burned nuclear fire 2 billion years ago.
One more thought: Oklo's reactors could only appear in a specific window of Earth's history. U-235 abundance fell from 3% to 0.72%, and the window closed forever. This suggests that many seemingly impossible things are not truly impossible — the timing just wasn't right. When the conditions align, even rock can burn with nuclear fire; when they scatter, even the best ore lies quietly underground.
Much like certain safety properties that only hold while model capability is insufficient. Once capability crosses a critical point, previously impossible behavior suddenly becomes inevitable. Oklo teaches us: critical points are not gradual changes but phase transitions. Between 0.72% and 3% lies not just a numerical difference, but the fundamental difference between "burns" and "doesn't."