Cowan and Reines: why two flashes were more convincing than one
Reactor antineutrinos became detectable through a correlated positron-and-neutron signature, followed by tests that challenged background explanations.
- 01Prompt
Positron signal starts the clock
- 02Delayed
Neutron capture can form a pair
- 03Control
Shift times to estimate accidentals
THE QUESTION
How can a detector distinguish a rare interaction from ordinary radioactivity?
Choose a reaction with two observable consequences
A reactor supplies electron antineutrinos from the beta decays of fission products. In inverse beta decay, an antineutrino interacts with a proton to produce a positron and a neutron. The positron slows and annihilates; the neutron slows and is captured later. The reaction creates a natural pair of signals rather than a single isolated flash.
Cowan, Reines and colleagues used this logic in the experiment reported in 1956. The Savannah River apparatus combined water targets containing cadmium with surrounding scintillation detectors. It was an arrangement designed around the products of an interaction, not an attempt to see the neutral incoming particle directly.
The delay is a piece of information
The prompt signal and the later neutron-capture signal have different physical causes. Capturing the neutron on cadmium yields gamma radiation that can be detected. The delay makes the pair recognizable while the energy selections and detector coincidences further constrain which records count as candidates.
A very narrow time window discards genuine captures that happen later. A very wide window admits unrelated pulses. There is no universally best window independent of the capture-time distribution, singles rates, energy selections and the question being optimized. This tradeoff is the central task of the accompanying Lab.
Sources: [2]
Accidental pairs also occur
Suppose unrelated prompt-like and delayed-like pulses arrive independently. A delayed pulse may happen to fall inside a prompt pulse’s window. Increasing the window increases this accidental population even though no additional neutrino interaction has occurred. At low occupancy, the accidental pair rate is approximately the product of the two singles rates and the window width.
The Lab generates a single stream of simulated pulses and then applies different windows to that same record. Candidate counts and accidental counts therefore come from actual selected pairs. A time-shifted control pairs records far apart in time to estimate the independent background. Simulation truth is available for learning, but a real experiment cannot ask each pulse whether it was signal.
Challenge the interpretation
The original evidence did more than count double pulses. The authors checked signal energies, the delay distribution and the dependence on reactor operation, and investigated competing radiation backgrounds. If the rate follows the reactor, that helps—but reactor-correlated neutrons and gamma rays must still be considered.
A useful way to read the paper is to ask what each check excludes. Timing challenges random coincidences. Energy selections test reaction products. Shielding and dedicated background measurements address particles that could imitate the signature. No single test does all of the work. The argument becomes strong because different alternatives fail different checks.
Use the Lab as a controlled comparison
Acquire a record with the reactor on. Sweep the coincidence window without resampling; distinguish improved signal acceptance from increased accidental acceptance. Next acquire a fresh reactor-off record. Finally increase the singles background and compare the candidate count with the time-shifted estimate. A rising count alone can be the wrong success metric.
The model uses illustrative rates, a single exponential capture lifetime, perfect pulse-class labels and no dead time. It is not a reconstruction of the 1956 dataset and does not represent the full instrument response of modern reactor experiments. Its purpose is to make the logic of a selection and its control sample inspectable.
Sources: [2]
Try the measurement
Primary sources & revision
- Cowan, Reines, Harrison, Kruse & McGuire · Detection of the Free Neutrino (1956)
- Reines et al. · Detection of the Free Antineutrino (1960)
First published and source-checked on 9 October 2026. Original explanatory text and diagrams by Sandbox Physics. Illustrations are schematic; no experimental event records are reproduced here. This is an independent educational publication, not an official Nobel or experiment collaboration publication.