Sterile neutrinos: a fourth kind?
A sterile neutrino would ignore even the weak force and reveal itself only by mixing with the known three. For thirty years puzzling anomalies pointed to one; the latest experiments have closed most of the doors, but not all.
Beginner
A neutrino that ignores almost everything
Ordinary neutrinos are famous for passing through matter almost untouched. Still, they do feel one force besides gravity: the weak interaction, which is how detectors catch them now and then. A sterile neutrino would not even feel that. It would respond only to gravity, so no detector could ever register it directly. The word "sterile" was used in this sense by Bruno Pontecorvo in 1967.
Counting experiments do not rule such a particle out. In 1989 measurements at the LEP collider at CERN showed that exactly three kinds of light neutrino couple to the Z boson, the carrier of the neutral weak force. A sterile neutrino does not couple to the Z, so it would not show up in that count of the number of light neutrinos.
How an invisible particle could be noticed
The trick is neutrino oscillation, the topic of its own deep dive. Ordinary neutrinos switch identity as they travel. If a fourth, sterile kind exists and mixes a little with the others, some neutrinos would spend part of their journey in the invisible state. A detector would then count slightly fewer neutrinos than expected, and the shortfall would rise and fall with distance. Think of a group of friends walking down a street where one of them now and then steps behind a curtain: the group seems to shrink and grow again as you watch.
Puzzling hints
Over three decades several experiments reported numbers that did not fit the three-neutrino picture.
- In the 1990s the LSND experiment at Los Alamos saw more electron antineutrinos than expected in a beam that should have contained almost none.
- Later the MiniBooNE experiment at Fermilab found too many events that looked like electron neutrinos.
- Detectors close to nuclear reactors counted about 6 % fewer antineutrinos than calculations predicted: the reactor anomaly.
- Gallium detectors exposed to strong radioactive sources produced less germanium than expected. The BEST experiment confirmed this deficit of about 20 % in 2022: the gallium anomaly.
Each of these hints could be explained by a sterile neutrino roughly as heavy as one electronvolt, which would make neutrinos oscillate over distances of metres rather than kilometres.
Checking the hints
Reactor experiments such as STEREO, PROSPECT and DANSS then placed detectors at several distances from a reactor core. A sterile neutrino would make the energy spectrum change with distance; none was seen. The reactor deficit now looks mostly like a flaw in the predicted reactor output. A claimed signal from the Russian Neutrino-4 experiment was contradicted by PROSPECT and by the tritium experiment KATRIN.
At Fermilab the MicroBooNE detector, which can tell electrons from photons much better than MiniBooNE, found no excess of electron neutrinos of the needed size. In 2025 it used two different neutrino beams and ruled out the simple single-sterile explanation of LSND and MiniBooNE. The IceCube observatory at the South Pole also found no clear signal.
Heavier relatives
Sterile neutrinos need not be light. Ones a few thousand times heavier could make up the dark matter; a faint X-ray line reported in 2014 was suggested as their decay signal, but the XRISM satellite did not see it. Much heavier ones, called heavy neutral leptons, could explain why ordinary neutrinos are so light, the topic of the deep dive on where neutrino masses come from.
Where things stand
The simplest picture, one light sterile neutrino explaining all hints, is now strongly disfavoured. Yet the MiniBooNE excess and the gallium deficit remain unexplained, and new detectors at Fermilab and an upgrade of KATRIN keep looking.
Quiz
Test your knowledge
4 questions on this level · Print as worksheet
Intermediate
Active and sterile states
The three known neutrinos are called active: they carry weak isospin and couple to W and Z bosons. A sterile neutrino is a fermion that is a singlet under all Standard Model gauge groups, typically identified with a right-handed neutrino. It can still mix with the active states. In the simplest extension, the 3+1 model, the PMNS matrix becomes a 4 × 4 matrix with new elements Ue4, Uμ4 and Uτ4, and a fourth mass state ν4 that is mostly sterile.
For Δm²41 ≈ 1 eV² the oscillation length is about 10 m for 4 MeV reactor antineutrinos and about 2.5 km at 1 GeV, so effects appear at baselines where the known three-flavour oscillations have not yet developed. This is why the anomalies below are called short-baseline anomalies.
The anomalies
- LSND (Los Alamos, 1993–1998) studied ν̄μ from muon decay at rest about 30 m from the target and found an excess of 87.9 ± 22.4 ± 6.0 ν̄e-like events, 3.8σ1. The similar KARMEN experiment saw no excess, leaving only part of the LSND region open.
- MiniBooNE (Fermilab Booster beam, 541 m) found 638 ± 133 electron-like events above background, concentrated at low energy, quoted at 4.8σ2. The collaboration combined this with LSND to 6.1σ.
- A 2011 re-evaluation of reactor fluxes gave an observed-to-predicted ratio of 0.943 ± 0.023 for all short-baseline reactor data3.
- Calibrations of the gallium solar experiments GALLEX and SAGE with ⁵¹Cr and ³⁷Ar sources gave capture rates below expectation. BEST, led by Vladimir Gavrin, placed a 3.4 MCi ⁵¹Cr source inside two nested gallium volumes and found ratios of 0.79 ± 0.05 and 0.77 ± 0.054.
Disappearance tests at reactors
Because reactor flux predictions are uncertain at the level of several per cent, newer experiments compare energy spectra at different distances, which cancels the flux model. STEREO at the ILL research reactor in Grenoble recorded 107,558 antineutrinos with six cells at about 10 m and rejected the reactor-anomaly hypothesis5. PROSPECT at the HFIR reactor in Oak Ridge excluded the best-fit point of Neutrino-4, which had claimed Δm²41 ≈ 7.3 eV² and sin²2θ ≈ 0.36 at 2.9σ, by more than 5σ. DANSS, on a movable platform below a commercial reactor at Kalinin, also sees no oscillation pattern. At the same time Daya Bay found that the deficit tracks the uranium-235 content of the fuel, pointing to the flux model rather than to oscillation.
KATRIN, built to measure the neutrino mass, also tests sterile neutrinos through a kink in the tritium β spectrum. Its analysis of 36 million electrons from 259 days excludes the Neutrino-4 claim and a large part of the region favoured by gallium and reactor data7.
Appearance tests
MicroBooNE, an 85-tonne active liquid-argon TPC at 470 m in the same beam as MiniBooNE, separates electrons from photons by their track topology and ionisation density. Its first analyses in 2021 found no excess of electron neutrinos matching MiniBooNE. In 2025 it combined the Booster beam with the off-axis NuMI beam and excluded the single-sterile explanation of LSND and MiniBooNE at 95 % CL6.
IceCube looks for a matter-enhanced resonance that a sterile neutrino would cause for TeV muon antineutrinos crossing the Earth. With 10.7 years of data it finds a best fit at Δm²41 = 3.5 eV², sin²2θ24 = 0.16, while the no-sterile hypothesis has a p-value of 3.1 %8, not significant.
Cosmology
An eV-scale sterile neutrino with the mixing required by the anomalies would be fully produced by oscillations before neutrino decoupling, raising Neff to about 4 and adding about 1 eV to the mass sum. Planck finds Neff = 2.99 ± 0.17 and Σmν well below 1 eV, so such a state would need extra new physics, such as new interactions, to hide from cosmology.
Quiz
Test your knowledge
4 questions on this level · Print as worksheet
Expert
The 3+1 framework and its consistency condition
For Δm²41 ≫ |Δm²31| and baselines where the three-flavour phases are negligible, the short-baseline probabilities depend only on the fourth column of the extended mixing matrix:
For small mixing this implies sin²2θμe ≈ ¼ sin²2θee sin²2θμμ. An appearance amplitude of order 10⁻³, as preferred by LSND and MiniBooNE, therefore requires both νe and νμ disappearance at the per-cent level. No such νμ disappearance has been seen. Global analyses, for example by Dentler, Schwetz and colleagues in 2018, found the appearance and disappearance data sets mutually incompatible at about 4.7σ within 3+1.
Electron disappearance: reactors and gallium
The reactor anomaly of 20113 rested on the conversion of measured ILL β spectra into antineutrino spectra. Two developments undermined it. First, Daya Bay's fuel-evolution analysis showed that the deficit is concentrated in ²³⁵U, while ²³⁹Pu agrees with prediction, which oscillation cannot produce. Second, a new measurement of the ratio of ²³⁵U and ²³⁹Pu β spectra at the Kurchatov Institute lowered the predicted ²³⁵U flux by several per cent, removing most of the deficit; the 5 MeV bump shows the spectral shape was mis-modelled too. Flux-independent ratio measurements (NEOS, DANSS, STEREO5, PROSPECT) now exclude the original reactor best fit. PROSPECT-I also excludes all gallium-preferred parameter space below 10 eV² at 95 % CL.
The gallium anomaly is harder to dismiss. BEST measured Rin = 0.79 ± 0.05 and Rout = 0.77 ± 0.05, with an outer-to-inner ratio of 0.97 ± 0.074, so no distance dependence is visible. A 3+1 interpretation needs sin²2θee of order 0.3 at Δm²41 ≳ 1 eV², in conflict with reactor ratio data, solar-neutrino constraints on |Ue4|² and KATRIN. KATRIN's 259-day analysis of 36 million electrons covers Δm²41 from a fraction of an eV² to several hundred eV² and disfavours the combined BEST–GALLEX–SAGE best fit7. Re-examinations of the ⁷¹Ga → ⁷¹Ge capture cross-section, including transitions to excited states, and of the ⁷¹Ge half-life have so far not removed the deficit.
Appearance after MicroBooNE
MicroBooNE's 2021 analyses found no νe excess matching the MiniBooNE template, and a separate analysis disfavoured a doubled rate of single photons from Δ(1232) radiative decay. A single Booster-beam measurement suffers from a degeneracy: intrinsic νe disappearance can compensate νμ → νe appearance. The 2025 analysis breaks it by adding NuMI data, whose νe fraction and energy spectrum differ, and excludes the 3+1 regions allowed by LSND and MiniBooNE at 95 % CL; a significant part of the gallium region is also excluded6. The MiniBooNE excess itself remains: proposed explanations include mis-modelled photon backgrounds and new particles decaying to e⁺e⁻ pairs or photons, which a Cherenkov detector cannot distinguish from electrons.
The Short-Baseline Neutrino programme at Fermilab adds SBND (112 t active, 110 m, data since 2024) and ICARUS (476 t, 600 m) to MicroBooNE's position at 470 m. Identical technology at three baselines cancels most flux and cross-section uncertainties.
Muon disappearance and IceCube
Because sterile states lack the neutral-current matter potential, TeV ν̄μ crossing the Earth's core would undergo an MSW-like resonance for Δm²41 ~ 1 eV², producing a large, zenith-dependent deficit. IceCube's 10.7-year analysis of 0.5–100 TeV events finds a best fit at Δm²41 = 3.5 eV², sin²2θ24 = 0.16, with the null hypothesis at p = 3.1 %8. This is not regarded as evidence.
keV sterile neutrinos as dark matter
A sterile neutrino with mass ms of a few keV and mixing sin²2θ ≲ 10⁻¹⁰ would be long-lived enough to be dark matter. Its radiative decay νs → ν γ produces an X-ray line at E = ms/2 with rate Γ ≈ 1.38 × 10⁻²⁹ s⁻¹ (sin²2θ / 10⁻⁷)(ms / 1 keV)⁵. Non-resonant production through oscillations (Dodelson–Widrow) is excluded by the combination of X-ray limits and small-scale structure; resonant production with a primordial lepton asymmetry (Shi–Fuller) survives in a narrow window. The 3.5 keV line reported in 2014 by two groups would correspond to ms ≈ 7.1 keV. XRISM's stacked spectrum of ten clusters (3.75 Ms) shows no line and sets a 3σ limit of about 1.0 × 10⁻²⁷ s⁻¹ on the decay rate, still about five times above the rate implied by the original XMM-Newton detection.
In the laboratory, a keV state would leave a kink in the tritium spectrum at E₀ − ms. KATRIN's TRISTAN upgrade, a silicon-drift-detector array that records the full spectrum and is planned for 2026, targets mixing near |Ue4|² ~ 10⁻⁶ for masses up to about 18 keV, far above cosmologically relevant values.
Heavy neutral leptons
Sterile states from MeV to TeV masses, heavy neutral leptons, are searched for in meson and W decays through their mixing |Uα|². Limits come from beam-dump and kaon experiments such as NA62, from B factories and from displaced-vertex searches at the LHC. The SHiP experiment, approved by CERN in 2024, will extend the reach for GeV-scale masses by several orders of magnitude. The νMSM of Asaka and Shaposhnikov combines one keV dark-matter state with two GeV states that generate neutrino masses and the baryon asymmetry, linking to the deep dive on where neutrino masses come from.
Quiz
Test your knowledge
4 questions on this level · Print as worksheet
Open questions
- What produces the MiniBooNE low-energy excess, if not a sterile neutrino: photon backgrounds, new particles decaying to e⁺e⁻ pairs, or something else?
- Is the gallium deficit due to nuclear physics, to the experiments themselves, or to new physics that does not look like simple 3+1 oscillation?
- Do sterile neutrinos exist at any mass scale, from keV dark matter to heavy neutral leptons that could explain neutrino masses?
- Could extended models, with several sterile states, decays or new interactions, still reconcile some anomalies with appearance, disappearance and cosmological data?
Terms to explore
- Sterile neutrino
- Short-baseline anomalies
- LSND
- MiniBooNE
- MicroBooNE
- Reactor antineutrino anomaly
- Gallium anomaly
- BEST (Baksan Experiment on Sterile Transitions)
- STEREO
- PROSPECT
- DANSS
- Neutrino-4
- KATRIN
- Sterile neutrino dark matter
- Heavy neutral lepton
People
- Bruno Pontecorvo
- William C. Louis
- Janet M. Conrad
- Bonnie T. Fleming
- Vladimir N. Gavrin (Vladimir Nikolaevich Gavrin)
- Thomas Schwetz
- Susanne Mertens
- Kathrin Valerius
- Mikhail Shaposhnikov
- Francis Halzen
On the timeline
- 1989LEP shows there are three light neutrinos
- 1995LSND reports an unexpected oscillation signal
- 2007MiniBooNE tests LSND and finds a low-energy excess
- 2018MiniBooNE's excess grows to 4.8σ
- 2021MicroBooNE finds no electron-neutrino excess
- 2022BEST strengthens the gallium anomaly
- 2010IceCube is completed at the South Pole
References
- A. Aguilar et al. (LSND), “Evidence for neutrino oscillations from the observation of ν̄e appearance in a ν̄μ beam”, Phys. Rev. D 64, 112007 (2001), doi:10.1103/PhysRevD.64.112007
- A. A. Aguilar-Arevalo et al. (MiniBooNE), “Updated MiniBooNE neutrino oscillation results with increased data and new background studies”, Phys. Rev. D 103, 052002 (2021), arXiv:2006.16883, doi:10.1103/PhysRevD.103.052002
- G. Mention et al., “The reactor antineutrino anomaly”, Phys. Rev. D 83, 073006 (2011), doi:10.1103/PhysRevD.83.073006
- V. V. Barinov et al. (BEST), “Results from the Baksan Experiment on Sterile Transitions (BEST)”, Phys. Rev. Lett. 128, 232501 (2022), arXiv:2109.11482, doi:10.1103/PhysRevLett.128.232501
- H. Almazán et al. (STEREO), “STEREO neutrino spectrum of ²³⁵U fission rejects sterile neutrino hypothesis”, Nature 613, 257 (2023), doi:10.1038/s41586-022-05568-2
- MicroBooNE Collaboration, “Search for light sterile neutrinos with two neutrino beams at MicroBooNE”, Nature 648, 64 (2025), arXiv:2512.07159, doi:10.1038/s41586-025-09757-7
- KATRIN Collaboration, “Sterile-neutrino search based on 259 days of KATRIN data”, Nature 648, 70 (2025), arXiv:2503.18667, doi:10.1038/s41586-025-09739-9
- R. Abbasi et al. (IceCube), “Search for an eV-scale sterile neutrino using improved high-energy νμ event reconstruction in IceCube”, Phys. Rev. Lett. 133, 201804 (2024), arXiv:2405.08070, doi:10.1103/PhysRevLett.133.201804
Last updated: · Editorial standard
Cite this page
Text
Neutrino Atlas (2026). “Sterile neutrinos: a fourth kind?”. Retrieved 22 November 2026, from https://neutrino-atlas.com/learn/sterile-neutrinos/BibTeX
@misc{neutrinoatlas2026learnsterileneutrinos,
title = {Sterile neutrinos: a fourth kind?},
author = {{Neutrino Atlas}},
year = {2026},
url = {https://neutrino-atlas.com/learn/sterile-neutrinos/},
note = {Accessed 2026-11-22},
language = {english}
}