In 1939, Soviet physicist Arkady Migdal predicted that when an atomic nucleus is struck hard enough to recoil, the electron cloud can't follow. The nucleus moves; the electrons, bound to the old position, are momentarily left behind. If the recoil is sharp enough, one electron tears free entirely. Two particles emerge from one collision: the recoiling nucleus and the ejected electron, diverging from the same point.
Reader note (2026-11-19): Checked against the open full text of the Nature paper, 'Direct observation of the Migdal effect induced by neutron bombardment' (January 2026, DOI 10.1038/s41586-025-09918-8, PMC12804083). The abstract says the result of 5 standard deviations rests on 6 candidate events selected out of almost 10^6 recorded events; the text has no hit for 800, so the essay's figure of 800,000 candidate events is not the paper's. The paper gives the ratio of the Migdal cross-section to the nuclear recoil cross-section as (4.9 +2.6 -1.9) x 10^-5, for nuclear recoils above 35 keVee and electron recoils of 5-10 keV, and says this is consistent with theoretical predictions; its estimate of the background is 0.59 +/- 0.40 expected delta electrons in the 5-10 keV range over the whole run. The detector is described as a gaseous pixel detector with a pixel array chip (83 micrometer pixels) and a deuterium-deuterium neutron source. The essay's 'three in ten million chance of coincidence' does not appear in the paper (no hit for 2.9, 3 x 10 or three in ten million); the paper states only the 5 standard deviations.
For 87 years this was theoretical. The signal was buried in noise — vanishingly rare, easily faked by background events, requiring a detector that could image individual atomic collisions with enough resolution to distinguish two tracks from one.
A team led by the University of the Chinese Academy of Sciences built the detector: a gas-based “atomic camera” combining a micro-pattern gas detector with a pixelated readout chip. They bombarded gas molecules with neutrons and sifted 800,000 candidate events. Six passed. Each showed two particle tracks — nucleus and electron — originating from the same point. The statistical confidence reached five sigma. Three in ten million chance of coincidence.
The finding matters for dark matter. Current dark matter detectors look for nuclear recoils — the tiny kick a dark matter particle gives an atomic nucleus when it collides. But light dark matter candidates produce recoils below the detection threshold. The nucleus moves, but too faintly to see. The Migdal effect offers a bypass. The nuclear recoil may be invisible, but the electron it ejects is not. Zheng Yangheng, one of the researchers: “With the Migdal effect, once an electron is ejected, our detector can, in theory, capture 100% of its energy.”
The atom's failure to stay coherent becomes the instrument. The electron can't keep up with the nucleus, and that lag — the atom's own internal delay — converts an undetectable recoil into a detectable electron. The weakness in atomic binding is the strength of the measurement. An 87-year-old prediction about what atoms cannot do becomes the tool for finding what we cannot see.