History

Every known atomic nucleus

—

Every atomic nucleus anyone has made or found — 3,386 of them — one square each, neutrons across and protons up. Press play to fill it in the order they were discovered. It does not grow outward from the middle: for thirty years everything found sits in the top corner, among the natural radioelements that announce themselves by decaying. The white spine of stable nuclei arrives almost at once — 7 of the 244 known in 1919, 133 by 1930 — once a mass spectrograph could weigh atoms one at a time. The proton-rich edge runs in flat ledges at 28, 50 and 82 protons — the closed shells. Switch the colour to half-life: of the 2,963 ever timed, 1,044 last under a second. Tap or hover any square.

How it comes apart

66 with no discovery year

Data & source

Data: IAEA Nuclear Data Section — Livechart of Nuclides API, the evaluated ENSDF/NUBASE ground-state file, served publicly and without a key · fetched —. The underlying evaluation is produced by the international evaluation network coordinated by the National Nuclear Data Center at Brookhaven; attribution is to the IAEA Nuclear Data Section and ENSDF. Scope: every ground state in the current evaluation — 3,386 of them — not a sample. One row per nuclide; excited isomers are not drawn. “Discovery” means the year of first reported observation, and the evaluation leaves it blank on 66 nuclides: those are drawn as hollow outlines through the whole pass and are never counted in the year total, rather than being silently dated. The earliest date in the file is 1851, on Ac-227 — forty-five years before radioactivity was discovered at all — so the timeline opens at 1896 with that one entry already on the chart. For several superheavy elements the first claim was contested for years, so the closing frames are a tidier history than the one that happened. 244 nuclides are marked stable, which means no decay has ever been observed rather than that none is possible; eight of those are watched as double-beta-decay candidates. 179 have no measured half-life at all — mostly unbound states out at the ragged edges — and they colour as “never measured” instead of being guessed. Where the evaluation gives a limit rather than a measurement (“> 1.6 × 10²⁵ y”), the tooltip keeps the limit. The file is a snapshot that keeps growing; this one stops at 2020. The table below counts by element rather than listing all 3,386 rows.