Radiocarbon Dating Explained: How Carbon-14 Works

A gram of living carbon emits 13.56 beta particles a minute, and the count fades after death. Here is why radiocarbon years are not calendar years.

Table of Contents

The Clock Starts When You Stop Eating

A gram of carbon taken from anything alive right now emits about 13.56 beta particles a minute. That is the measured specific activity of modern carbon, and it exists because roughly 1.2 out of every trillion carbon atoms in the biosphere is carbon-14 rather than ordinary carbon-12. You are about 18% carbon by mass, so a 70 kg adult contains some 12.6 kg of it, which works out to around 2,800 carbon-14 decays every second inside your body. You are faintly, continuously radioactive, like every tree, fish and loaf of bread.

While you are alive the ratio holds steady, because you keep eating. Carbon flows in from the biosphere, which is replenished from the atmosphere, which is replenished by cosmic rays. The instant that exchange stops — death, a branch cut, a seed charred in a fire — the top-up ends and the carbon-14 begins to disappear with nothing behind it.

That is the whole method. Among radioactive dating techniques, radiocarbon is unusual: the event it times is not the formation of a rock but the end of a life. Willard Libby worked it out at Chicago in the late 1940s, published the first dates with James Arnold in Science in December 1949 after testing them against Egyptian tomb material of known age, and took the 1960 Nobel Prize in Chemistry.

Cosmic Rays, Nitrogen, and a Steady Leak

Carbon-14 is not primordial. It is manufactured continuously, high up, by cosmic rays smashing into air molecules and spraying out neutrons. A thermalised neutron meeting a nitrogen nucleus does something unusually tidy:

n + ¹⁴N → ¹⁴C + p

The neutron goes in, a proton comes out, and the nucleus keeps its mass number but loses a proton. Nitrogen is 78% of the air, which is why this channel dominates. Estimates of the global production rate have moved as cosmic-ray models improved: older calculations gave about 2.5 atoms per square centimetre of Earth’s surface per second, newer Monte Carlo models closer to 1.6 to 1.9. Either way it amounts to a few kilograms a year — roughly 7.5 kg, or 3.2 × 10²⁶ atoms — for the entire planet.

Those atoms oxidise to carbon dioxide within hours and join the circulation of the atmosphere, mixing between hemispheres in a year or two and entering plants through photosynthesis. Production roughly balances decay, which is what keeps the concentration near constant on human timescales. Roughly. That word does a great deal of work here.

The Half-Life Nobody Quite Agrees On

Carbon-14 decays by beta emission back to nitrogen-14, and the surviving fraction follows the ordinary exponential decay law:

f = e^(−t/τ), so t = τ·ln(1/f)

where τ is the mean life, equal to the half-life divided by ln 2. The trouble is which half-life. The current evaluated nuclear-data value is 5,700 ± 30 years. The Cambridge value of 5,730 ± 40 years, agreed in 1962, is still what most textbooks print. Libby’s own 1949 measurement gave 5,568 ± 30 years.

Radiocarbon laboratories still use Libby’s number. By a convention formalised in 1977, a published conventional radiocarbon age in years BP — before 1950 — is calculated with the 5,568-year half-life and a fixed assumed starting activity, which gives a mean life of exactly 8,033 years. Nobody thinks that half-life is correct. Keeping it means a date measured in 1958 and a date measured last week can be compared directly, and the 2.4% systematic error it introduces is removed later by calibration. It is one of the more honest fudges in science: a wrong constant, used deliberately and consistently, with the correction applied at the end.

The same exponential sets the ceiling. After nine half-lives only 0.2% of the carbon-14 survives, about the size of the stray modern carbon no laboratory keeps entirely out of its reagents and targets. Contamination bites harder than counting statistics: 1% of modern carbon in a genuinely 50,000-year-old sample makes it measure as about 36,000 years. That is why 50,000 years is the working limit and why dates near it get argued over.

The decay curve, and why it stops being useful near 50,000 years Conventional radiocarbon age: t = 8033 · ln(1/f), from the 5,568-year Libby half-life used by convention. Panel 1 — surviving ¹⁴C fraction f ← one half-life per dashed line 100% 75% 50% 25% 0 practical limit about 50,000 yr: 0.2% left, and lab background is the same size Shroud of Turin linen, 691 ± 31 BP (1988) Ötzi the Iceman, 4,550 ± 19 BP Lascaux charcoal, 17,190 ± 140 BP 0 10,000 20,000 30,000 40,000 50,000 60,000 conventional radiocarbon age (years BP) Panel 2 — how big the error bar gets 10,000 yr 1,000 100 10 Modelled as 0.3% precision on the ¹⁴C/¹²C ratio plus a blank equivalent to 0.05% modern carbon. Real laboratory limits differ; the shape of the blow-up does not. 0 10,000 20,000 30,000 40,000 50,000 60,000 BP ¹⁴C remaining 56.76% 7.70 decays/min per gram conventional age 4,549 BP ± 25 yr (1σ, modelled) same f, 5,700-yr half-life 4,657 yr 2.4% older than conventional half-lives elapsed 0.82 5,568 yr each, by convention
56.76%
Drag the slider from Oetzi towards the right-hand wall and watch the blue error bar in panel 1 swell from invisible to wider than a millennium: the method does not fail suddenly, it dissolves. Ages shown are conventional radiocarbon years, not calendar dates.

Radiocarbon Years Are Not Calendar Years

The method assumes it knows how much carbon-14 was in the air when the organism died. It does not. Atmospheric concentration drifts with solar activity, which modulates the cosmic-ray flux, with Earth’s magnetic field, which deflects it, and with ocean circulation, which buries and releases old carbon. A raw measurement therefore has to be mapped onto real time using material that is both datable and independently countable.

That curve is IntCal20, published in 2020 and agreed internationally, running from the present to 55,000 calibrated years before present. Out to about 13,900 years it is built on tree rings, counted one by one; beyond that it is stitched together from floating tree-ring sequences, speleothems, corals, and lake and marine sediments. The curve wiggles and in places flattens: a stretch of the first millennium BC is so flat that one measurement can be consistent with two or three separate calendar windows, which is why a properly reported date is a probability distribution over intervals rather than a number.

The Twentieth Century Broke the Baseline

Atmospheric nuclear testing in the 1950s and early 1960s roughly doubled the carbon-14 content of the northern atmosphere, peaking in 1963 in the north and in late 1965 in the south. The excess has since drained into the oceans and biosphere with a decline time of about 16 years, and the spike is a forensic gift: because the curve is steep, tissue grown in those decades can be dated to within a year or two, which is used on ivory, wine and human remains.

Pushing the other way is the Suess effect. Fossil fuels have been underground long enough that they contain no carbon-14 at all, so burning them dilutes the atmospheric ratio. The dilution currently adds roughly 30 years of apparent age to the atmosphere every year. Heather Graven calculated in 2015 that under a high-emissions path, fresh plant material in 2050 could carry the radiocarbon signature of something a thousand years old, and two thousand years old by 2100. One of the stranger consequences of a changing carbon cycle is that we are quietly breaking a dating method.

Counting Atoms Instead of Waiting for Decays

Libby counted decays. A gram of modern carbon gives 13.56 counts a minute, so a sample ten half-lives old gives about one count every forty minutes, and the only way to get statistics is more carbon and more time: classical beta counting wants grams of charcoal, or up to a hundred grams of bone, and days in a shielded counter.

Accelerator mass spectrometry does not wait. It ionises the sample, accelerates the ions through a small tandem accelerator, strips their electrons to destroy the molecular interferences that would otherwise swamp mass 14, and counts individual carbon-14 atoms in a detector. Because it counts atoms rather than the few that happen to decay while you watch, it needs about a milligram of carbon instead of ten grams, and hours instead of days. That is what made it possible to date a thread, a single seed, or one line of pigment.

What the Dates Actually Settled

In 1988 laboratories in Arizona, Oxford and Zurich measured samples cut from the Shroud of Turin. The unweighted mean was 691 ± 31 BP, calibrating to AD 1260 to 1390 with at least 95% confidence, published in Nature in 1989. The result has been contested ever since on sampling and contamination grounds, and the argument is still live.

Ötzi, the body that melted out of an Alpine glacier in 1991, gave a combined age of 4,550 ± 19 BP, calibrating to roughly 3370 to 3100 BC. Lascaux has a longer history: Libby himself ran a charcoal sample from the Shaft in 1951 and got 15,516 ± 900 BP, and André Glory’s later Passage samples gave 17,190 ± 140 BP. The paintings are now usually placed between 17,000 and 22,000 years old, which is a reminder that a radiocarbon date tells you the age of the charcoal, not necessarily of the hand that drew with it.

All of it rests on one number: how much of a trillionth is left. It is a strange foundation for the chronology of human history — one atom in 830 billion, and going away as you read.

Frequently Asked Questions

How does radiocarbon dating work?

Cosmic rays striking the upper atmosphere produce neutrons that convert nitrogen into carbon-14 through the reaction n + ¹⁴N → ¹⁴C + p. That carbon-14 oxidises to carbon dioxide, mixes through the atmosphere and enters every living thing by photosynthesis and eating, so all living tissue holds roughly 1.2 carbon-14 atoms per trillion carbon atoms, giving a measured activity of about 13.56 decays per minute per gram of carbon. While an organism is alive the ratio stays topped up. When it dies the exchange stops and the carbon-14 decays away with nothing replacing it. Measuring how much is left therefore measures time since death: the age is t = τ·ln(1/f), where f is the surviving fraction and τ is the mean life. Laboratories report conventional radiocarbon ages using τ = 8,033 years, then convert them to calendar dates with a calibration curve.

What is the half-life of carbon-14?

There are three numbers in circulation and it matters which one you mean. The current evaluated nuclear-data value is 5,700 ± 30 years. The older Cambridge value of 5,730 ± 40 years, adopted in 1962, is still very widely quoted in textbooks. Willard Libby's original 1949 measurement gave 5,568 ± 30 years, and that is the value radiocarbon laboratories still use by international convention when they publish a conventional radiocarbon age in years BP, where BP means before 1950. They do not use it because anyone believes it; they use it so that seventy years of published dates stay comparable, and the resulting 2.4 per cent error is absorbed when the date is calibrated. A half-life of 5,568 years gives a mean life of 8,033 years; 5,700 years gives 8,223.

Why can radiocarbon dating only reach about 50,000 years?

Because the signal runs out. After 50,000 years only about 0.2 per cent of the original carbon-14 remains, which is roughly nine half-lives, and what is left is comparable to the laboratory's own background: stray modern carbon in reagents, the sample holder, the graphite target. Contamination is the harder wall. One per cent of modern carbon added to a genuinely 50,000-year-old sample makes it measure as about 36,000 years old, so any date near the limit depends entirely on chemical pretreatment that may not be perfect. Accelerator mass spectrometry with very clean chemistry can reach 55,000 years on favourable material, and the IntCal20 calibration curve now extends to 55,000 calibrated years before present, but uncertainties grow from a few decades at young ages to thousands of years out there. Beyond that, other methods take over.

What is the difference between radiocarbon years and calendar years?

A radiocarbon year is a measurement, not a date. Converting one into the other requires a calibration curve, because the amount of carbon-14 in the atmosphere has never been constant: it varies with solar activity, with Earth's magnetic field, and with how much old carbon the oceans bring up. A raw result is reported as a conventional radiocarbon age in years BP, calculated with the fixed 5,568-year Libby half-life and a fixed assumed starting concentration. The internationally agreed curve, IntCal20, published in 2020, maps those measurements onto calendar years out to 55,000 cal BP, using tree rings as far back as about 13,900 years and then floating tree-ring sequences, speleothems, corals and lake and marine sediments. Because the curve has flat stretches, a single radiocarbon measurement can sometimes correspond to two or three separate calendar intervals.

Read Next