Dug 0.00 mLayer 1Soil removed 0.0 m³Finds exposed 0select a findFPS 0

Drag to orbit, wheel or pinch to zoom. The walls (baulks) show the layers; the red ring marks the selected find. Dig with the slider or the 5 cm button, pick an exposed find, and count its carbon-14 decays.

Excavation Trench: Stratigraphy, Radiocarbon Dating and Contamination

Dig through five layers, pick a find and count its carbon-14 decays. Does the date agree with the layer it came from?

Excavation

Include a later pit that cuts through the layers (intrusion)

Finds (exposed ones can be dated)

Counting

Select an exposed find, then press Count decays. The counter runs the chosen hours of simulated time.

Live readouts

Running age estimate while counting

Repeat the measurement 600 times

Age against depth

Harris matrix (context order)

Missions

Predict: how many hours must a 1 g sample that is 3000 radiocarbon years old be counted (background 0.02 cps) to reach a 1-sigma age uncertainty of 50 years?

Hint: the uncertainty falls as the square root of the number of counts.

Experiments

Each button measures the selected find 1500 times per setting with the current sliders and reports the mean and spread of the ages.

Independent numerical checks

Controls

Drag / one finger: orbit; wheel / pinch: zoom; three camera views.
Dig depth is a slider; a find becomes datable once the trowel has uncovered it.
Presets set the counting conditions; the contamination slider models modern carbon (rootlets, handling, cleaning chemicals) mixed into the sample.

The science

Law of superposition. In undisturbed deposits the lowest layer is the oldest. The trench here has five contexts: topsoil, hillwash (colluvium), a burnt destruction layer, an occupation floor and sterile natural subsoil. Ages are modelled as a linear deposition rate inside each layer (350 y at 0.25 m, 1700 y at 0.65 m, 1850 y at 0.75 m, 3200 y at 1.05 m). These numbers are invented for teaching, not from a real site.

Radioactive decay. Carbon-14 decays as N(t) = N₀ e^-t/τ with mean life τ = t_1/2/ln 2, half-life 5730 y (mean life 8267 y). Living things keep the modern 14C fraction; after death it only decays. The modern activity is about 0.226 Bq per gram of carbon.

Conventional radiocarbon age. Laboratories report t = -8033 ln(A/A_mod) years BP, using the historic Libby half-life 5568 y. To get closer to true years multiply by 5730/5568 = 1.029, then apply a calibration curve (IntCal) for past changes in atmospheric 14C. This simulation applies the half-life factor but has no calibration curve, so real calendar ages can differ by decades to centuries.

Counting statistics. Decays are random: the counts N in time T are Poisson distributed with standard deviation √N. The net sample rate is the sample counts minus the background counted equally long, so σ_t = τ_L √(r_s + 2 b) T / (r_s T) with sample rate r_s and background b. Four times the counting time halves the uncertainty. A detection efficiency of 0.9 and 0.02 cps background are illustrative values.

Contamination. A mass fraction c of modern carbon raises the measured fraction to R = c + (1-c) e^{-t/τ}. The apparent age can never exceed -τ_L ln c: 1 % contamination makes everything older than about 37 000 y look 37 000 y old, and 5 % shifts a 3000-year-old sample by roughly 400 years.

Intrusions. A pit dug later and filled with younger soil puts a young find physically below older ones. The stratigraphic order alone then misleads; a radiocarbon date that is far younger than its layer is the clue. The Harris matrix records that the pit cut through contexts 2 to 4 after they formed.

Simplifications. One mean life, no isotopic fractionation correction, no marine or old-wood effects, no calibration curve, perfect sample cleaning except for the contamination you add, counting efficiency and background constant. Ages are for the lesson only. Assets built with Blender; scene by Three.js.