An anomalous measurement from a nuclear reactor triggered a three-year campaign to find an elusive particle called the sterile neutrino. The search shows definitively that sterile neutrinos don’t exist — but the anomaly persists.

Neutrinos are among the most abundant elementary particles in the Universe, but they have zero electric charge and interact only weakly with matter, so are difficult to detect in experiments. They have therefore been implicated as the reason behind some of the key gaps in our current understanding of the Universe. The long-held idea1 that there are only three types of neutrino was challenged in 19962 by evidence suggesting the possibility of a fourth type, called the sterile neutrino. Further support for this proposal came in 20113, when the total number (the flux) of antineutrinos — the antimatter counterpart of neutrinos — produced in a nuclear reactor differed significantly from that predicted. A dedicated search ensued. And now, in a paper in Nature, the STEREO collaboration4 confirms the flux anomaly, but reports that this discrepancy cannot be explained by the existence of a sterile neutrino.

In 1989, experiments1 on the Large Electron–Positron Collider (LEP) at CERN, Europe’s particle-physics laboratory near Geneva, Switzerland, determined precisely that there were three types (or ‘flavours’, in particle-physics terms) of neutrino. The three confirmed flavours are the electron neutrino, the muon neutrino and the tau neutrino. Neutrinos are generated when cosmic rays interact with Earth’s atmosphere, and also through nuclear fusion occurring in the Sun’s core. Experiments designed to detect these atmospheric5 and solar6 neutrinos established the curious fact that neutrinos oscillate — they change spontaneously from one flavour to another as they travel.

Neutrino oscillation can occur only if neutrinos have mass, and their mass is difficult to measure. An electron neutrino (or any other flavour) is a quantum mixture of three states that have different masses. When this neutrino moves through space, quantum interference between the three states leads to the periodic flavour transformations that constitute neutrino oscillation.

Although scientists do not know the exact values of the neutrino masses, they can measure the differences between them; these are called ‘mass splittings’ and are proportional to the oscillation frequencies...

Electron antineutrinos are produced in abundance by nuclear-fission reactions in the cores of nuclear reactors. This type of antiparticle has been studied extensively, and it had a key role in the 1956 discovery of the neutrino7. To use reactor-generated electron antineutrinos to investigate neutrino oscillation, it is crucial to know both the flux of antineutrinos emitted in these nuclear reactions and how many antineutrinos are produced at particular energies (the energy spectrum). The precision with which these quantities are measured and theoretically predicted has improved markedly since the 1950s. And in 2011, it was found that the average antineutrino flux detected in these experiments was about 6% less than that predicted8.

One possible explanation8 for this reactor antineutrino anomaly is that some neutrinos morph into sterile neutrinos after they leave the reactor core. Dedicated experiments were designed to investigate this possibility by installing detectors close to the reactor, usually at a distance of around 10 metres, where an oscillation into sterile neutrinos might show up in the observed energy spectrum. However, these experiments made the picture only messier: some reported that such a signature was observed9,10, and others reported a negative result11,12...

...Because neutrinos are so difficult to detect, signals from sources other than the reactor can be many times more abundant than real neutrino signals if the detector is not well shielded. These false signatures are called background signals, and they can blur or distort the measurements. The authors were able to avoid many uncertainties by using their comparative measurement. The background signals were well controlled with relatively good shielding, and were measured when the reactor was switched off...

In addition to the flux deficit, the authors observed that the energy spectrum was distorted with respect to the model predictions — with a ‘bump’ between 5 and 6 megaelectronvolts, which was also detected in previous experiments13–15