Breathe in. Roughly one in every trillion carbon atoms you just took up is carbon-14, which is unstable. A few thousand of them are breaking apart inside you every second, and the supply never runs down, because you ate breakfast.
Two processes fix that number, and neither is happening in this room. Cosmic rays smash into the upper atmosphere and spray secondary neutrons downward. A neutron that finds a nitrogen-14 nucleus can knock out a proton and take its place, and the atom that walks away is carbon-14. Production peaks nine to fifteen kilometres up and toward the poles, where the geomagnetic field offers least shelter, at roughly two atoms per square centimetre of Earth’s surface per second. Meanwhile, everywhere, carbon-14 decays — shedding an electron, turning back into the nitrogen it came from, with a half-life near 5,700 years.
Begin with what that ratio is not: constant. Cosmic-ray flux rises and falls with the sun’s magnetic activity, and the field deflecting it drifts over millennia. Tree rings record the consequences ring by ring, including single-year spikes sharp enough to serve as calendar anchors — the largest in AD 774–5 and AD 993. The whole apparatus of calibration curves exists because the level moves. Radiocarbon dating is not spared that. It is built on top of it.
And the level is not really the atmosphere’s to set. Of the carbon circulating between air, life and sea, the ocean holds the overwhelming majority — some sixty times what the pre-industrial atmosphere held. Air, plants and the sea surface come to terms within decades; the deep ocean takes thousands of years. So carbon-14 is manufactured in a thin shell atop a small reservoir and destroyed mostly in an enormous one beneath — the air the narrow window between, holding a value neither rate quite controls.
The asymmetry shows up in dates. Shellfish alive at the same moment as a tree come out some four hundred radiocarbon years older, and over a thousand in polar seas, because the water they built their shells from last touched the sky centuries ago and has been decaying since. The sea does not carry the sky’s number. It carries a delayed, diluted copy.
The clearest demonstration was an accident. Atmospheric weapons testing nearly doubled Northern-Hemisphere carbon-14 by 1963. Then the test ban took effect and the level fell steeply, with a residence time near sixteen years. Almost none of that fall was decay — over sixteen years, decay removes about two parts in a thousand. The atoms did not disappear; they were absorbed into oceans and vegetation. The bomb spike is one of the finest measurements ever made of how fast air trades carbon with everything else, and says essentially nothing about half-lives.
Which is the real reason the method works. A living thing is plumbed into the reservoir: it eats, breathes, and holds the ambient ratio without effort. Death cuts the connection, nothing is topped up again, and the decay continues alone. So the age of a bone is not read off a clock inside it, but off the gap between the bone and a level still being maintained outside it. You are not timing the sample. You are measuring how far it has drifted from a standing balance it left.
Even the arithmetic carries a fossil of this. Laboratories still compute conventional radiocarbon ages from Libby’s original half-life of 5,568 years, though a better figure — 5,730, now nearer 5,700 — arrived within a decade. The wrong number is kept on purpose, so dates published across seventy years stay comparable; calibration absorbs the discrepancy.
We have shoved the balance twice in one century, in opposite directions. The bombs pushed it up, and that push drained away within decades. The other shove does not drain. Coal, oil and gas are old enough that every carbon-14 atom they once held is long gone, so burning them injects carbon that dilutes the ratio — the Suess effect, already some thirty parts per thousand below pre-industrial by the late twentieth century.
It ends somewhere strange. On a high-emissions path, modelling puts the air’s signature near a thousand-year-old atmosphere by 2050 and a two-thousand-year-old one by 2100. A plant grown in that air would date, honestly and correctly, to about AD 100. The isotope will not have changed. Neither will the half-life. We will simply have moved the level every date is measured against, and the clock will go on reading out perfectly, and be wrong.
The archive has met this shape twice. No. 84’s chemostat holds a population steady, but a person set the pump. No. 29’s ozone is made and destroyed by the same sunlight in the same layer — a balance that is at least local. This is neither. Nobody set either rate, and the rates do not happen in the same place. What every living thing carries is a number settled between a factory fifteen kilometres overhead and a graveyard four kilometres down.
≈
One loop I’m watching
Next: a few hundred metres below the surface of the Pacific there is a layer where oxygen nearly vanishes — not because anything sealed it off, but because sinking dead plankton is respired there faster than circulation can resupply it. The layer keeps a depth and a thickness for as long as anyone has measured it, while every molecule of water in it is only passing through.