Atmospheric neutrinos
Cosmic rays striking the upper atmosphere create neutrinos that reach every detector from all directions. Some have flown 15 kilometres, others have crossed the whole planet. This natural spread of distances made atmospheric neutrinos the first proof that neutrinos change flavour.
Beginner
Rain from cosmic collisions
Space is full of fast atomic nuclei, mostly protons, called cosmic rays. When one of them strikes a nitrogen or oxygen nucleus some 15 to 20 kilometres above the ground, it shatters the nucleus and starts an air shower, a cascade of short-lived particles. Among them are pions and kaons, which decay within a tiny fraction of a second. Their decay products include muons and neutrinos, and the muons in turn decay into more neutrinos. These are the atmospheric neutrinos. Typical energies are around a billion electronvolts and above, hundreds of times more than neutrinos from the Sun.
Two to one
The chain of decays follows simple bookkeeping. A pion decays into a muon and a muon neutrino. The muon then decays into an electron, an electron neutrino and another muon neutrino. So for every electron neutrino one expects about two muon neutrinos. This ratio can be predicted far more precisely than the total number of neutrinos, because many uncertainties cancel when one number is divided by another. That made it a sharp test. At higher energies many muons reach the ground before they decay, and the ratio becomes larger than two.
A natural experiment with every distance
Neutrinos cross the Earth almost unhindered, so a detector deep underground sees atmospheric neutrinos from every direction. Those from directly overhead have flown about 15 kilometres. Those from the horizon have travelled a few hundred kilometres. Those from directly below have crossed the whole planet, about 13,000 kilometres. Cosmic rays hit the Earth roughly equally from all sides, so without any change on the way, as many muon neutrinos should arrive from below as from above. Comparing the two is like timing identical runners, one on a short track and one on a marathon course: any difference must happen along the way.
The 1998 discovery
In the late 1980s the Kamiokande detector in Japan and IMB in the United States found too few muon neutrinos compared with electron neutrinos. Many physicists suspected a measurement error. Super-Kamiokande, a tank of 50,000 tonnes of ultra-pure water a kilometre under a mountain in Japan, settled the question. In June 1998 Takaaki Kajita presented the result: energetic muon neutrinos coming up through the Earth were only about half as numerous as those coming down, while electron neutrinos showed no such difference1. The missing muon neutrinos had turned into tau neutrinos on their long journey. This was the first compelling proof that neutrinos have mass, and Kajita shared the 2015 Nobel Prize in Physics for it. The wider story is told in the lesson “Neutrino oscillation”.
Large detectors in ice and sea
Besides Super-Kamiokande, the IceCube observatory at the South Pole studies atmospheric neutrinos with its densely instrumented inner part, DeepCore. In the Mediterranean, the ORCA detector of KM3NeT is being built off the French coast for the same purpose. These detectors now record tens of thousands of atmospheric neutrinos. One of their goals is to find out which neutrino is the heaviest, the mass ordering. Neutrinos that cross the dense mantle and core of the Earth are affected slightly differently depending on the answer.
Signal and nuisance
For astronomers searching for neutrinos from distant galaxies, atmospheric neutrinos are an unavoidable background, much like city lights for someone looking at faint stars. The same holds for searches for the decay of the proton and for the faint glow of neutrinos left by all past supernovae. Knowing the atmospheric flux precisely therefore matters far beyond the study of oscillations.
Quiz
Test your knowledge
4 questions on this level · Print as worksheet
Intermediate
From cosmic ray to neutrino
Primary cosmic rays are mostly protons, with some helium and heavier nuclei, and their spectrum falls steeply with energy. Collisions 15 to 20 km up produce mainly charged pions and, increasingly at higher energies, kaons. The decay chain π⁺ → μ⁺ νμ followed by muon decay μ⁺ → e⁺ νe ν̄μ, together with the charge-conjugate chain, gives (νμ + ν̄μ) : (νe + ν̄e) ≈ 2 : 1 below a few GeV. A 1 GeV muon travels about 6 km before decaying, so above a few GeV most muons reach the ground first and the ratio rises. Above about 10 GeV the neutrino spectrum falls roughly as E⁻³·⁷.
Flux calculations, such as the widely used one by Honda and collaborators2, predict absolute rates only to about 10–20 %, mainly because of uncertainties in hadron production. Flavour ratios and the up–down symmetry of the flux are known to a few percent, and these were the quantities that exposed oscillation.
Baseline from the zenith angle
The zenith angle fixes the flight path: about 15 km for neutrinos from directly above, a few hundred kilometres near the horizon and about 12,700 km for neutrinos from directly below. With energies from 0.1 GeV to beyond 100 GeV, the ratio L/E spans more than five orders of magnitude in a single detector. Above a few GeV the Earth's magnetic field hardly affects the primaries, so the flux from below should mirror the flux from above.
Kamiokande, IMB and Super-Kamiokande
In 1988 Kamiokande found that the ratio of muon-like to electron-like events was only about 60 % of the prediction, and IMB saw a similar deficit. Super-Kamiokande, with 50 kt of water and about 11,000 large photomultipliers, separates sharp muon rings from fuzzy electron showers by ring imaging. From a 33 kt·yr exposure (535 days) it found for multi-GeV muon-like events an up–down asymmetry (U − D)/(U + D) = −0.296 ± 0.048 (stat.) ± 0.01 (syst.), while electron-like events were symmetric1. A two-flavour νμ → ντ fit gave sin²2θ > 0.82 and 5 × 10⁻⁴ < Δm² < 6 × 10⁻³ eV² at 90 % confidence. Yoji Totsuka led the collaboration at the time.
In 2004 Super-Kamiokande saw the first dip in the event rate as a function of L/E, which excluded neutrino decay and decoherence as explanations. In 2018 it reported the statistical appearance of tau neutrinos in the up-going sample at 4.6σ3, direct evidence that the muon neutrinos turn mainly into tau neutrinos.
Ice and deep water
DeepCore, the dense infill of IceCube, lowers the energy threshold to about 10 GeV, enough to see the first oscillation minimum for up-going neutrinos near 25 GeV. With data from 2011 to 2019 it measured sin²θ23 = 0.51 ± 0.05 and Δm²32 = (2.41 ± 0.07) × 10⁻³ eV², assuming normal ordering4, comparable in precision with accelerator experiments. KM3NeT/ORCA, about 2,450 m deep off Toulon, will have 115 detection units. With only six of them it measured sin²θ23 = 0.51 (+0.04, −0.05) and Δm²31 ≈ 2.2 × 10⁻³ eV² from 5,828 events5.
Matter effects and the ordering
Up-going neutrinos cross the mantle and, at the steepest angles, the iron core. Coherent scattering on electrons, the Earth matter effect, resonantly enhances νμ ↔ νe mixing at energies of a few GeV: for neutrinos if the ordering is normal, for antineutrinos if it is inverted. Water and ice detectors cannot tell ν from ν̄ event by event, but neutrinos have a larger flux and roughly twice the cross-section, so a net effect survives. Super-Kamiokande's atmospheric data drive a mild preference for the normal ordering in global fits6; the dedicated lesson on the mass ordering compares this with other methods.
Quiz
Test your knowledge
4 questions on this level · Print as worksheet
Expert
Flux and flavour composition
Below a few GeV the conventional flux comes from π and μ decays in flight; for νμ the kaon contribution takes over above roughly 100 GeV. Two numbers control the flavour ratio:
The absolute normalisation carries 10–20 % uncertainty from the primary spectrum and hadron production2; ν/ν̄ ratios, flavour ratios and up–down ratios are predicted to a few percent. The prompt component from charm decays, with a harder spectrum near E⁻²·⁷ and r ≈ 1, has not yet been identified.
Geometry and observables
For vertically up-going neutrinos the first minimum falls near E ≈ 25 GeV, within reach of DeepCore; in Super-Kamiokande the multi-GeV up-going sample sits in the averaged regime, which produces the deficit of about one half. The 1998 asymmetry A = (U − D)/(U + D) = −0.296 ± 0.048 ± 0.01 for multi-GeV μ-like events1 corresponds to U/D ≈ 0.54. Sub-GeV electron-like events carry a small Δm²_21-driven excess proportional to (r cos²θ23 − 1), which vanishes for r = 2 and maximal mixing and therefore probes the octant of θ23.
Matter effects in the Earth
For Δm²31 > 0 the resonance enhances νμ → νe for neutrinos, otherwise for antineutrinos. Core-crossing trajectories add a parametric resonance at 3–5 GeV where mantle and core phases add coherently. The resulting νe excess and νμ distortions at cos θz < −0.4 and 3–10 GeV are the ordering signature. Because σ(ν) ≈ 2σ(ν̄) at GeV energies, charge-blind detectors retain a net sensitivity; magnetised calorimeters such as the planned INO-ICAL would separate μ⁻ from μ⁺ directly.
Current results
IceCube DeepCore, 2011–2019: sin²θ23 = 0.51 ± 0.05 and Δm²32 = (2.41 ± 0.07) × 10⁻³ eV² (normal ordering)4. KM3NeT/ORCA6, 433 kt·yr: sin²θ23 = 0.51 (+0.04, −0.05), Δm²31 = 2.18 (+0.25, −0.35) × 10⁻³ eV², inverted ordering disfavoured only at p = 0.255. Super-Kamiokande: ντ appearance at 4.6σ3. In NuFIT 6.0 the Super-Kamiokande atmospheric χ² shifts the inverted ordering to Δχ² = 6.1 and moves θ23 into the lower octant6. The IceCube Upgrade, deployed in 2025/26, and the growing ORCA array target the ordering, especially in combination with JUNO, whose reactor measurement constrains |Δm²31| independently of matter.
Atmospheric neutrinos as background
They set the irreducible background for proton decay searches in water Cherenkov detectors, close the DSNB window above about 30 MeV and dominate the IceCube event rate below roughly 100 TeV, where the self-veto of down-going showers helps. For dark matter detectors they shape the high-mass edge of the neutrino fog.
Quiz
Test your knowledge
4 questions on this level · Print as worksheet
Open questions
Atmospheric data alone do not yet settle the mass ordering or the octant of θ23; the hints from Super-Kamiokande, DeepCore and ORCA point in slightly different directions and need larger exposures and better detector calibration.
The absolute flux is still uncertain at the 10–20 % level, limited by hadron production in air showers, and the prompt flux from charmed mesons has never been measured.
Terms to explore
- Atmospheric neutrinos
- Cosmic rays
- Cosmic-ray air showers
- Zenith angle
- Super-Kamiokande
- DeepCore & IceCube upgrade
- KM3NeT
- Earth matter effects
- Neutrino mass ordering
- Tau neutrino appearance
People
On the timeline
- 1965First detection of atmospheric neutrinos
- 1988Kamiokande finds an atmospheric neutrino anomaly
- 1998Super-Kamiokande discovers neutrino oscillation
- 2004Super-K sees the oscillation dip in L/E
- 2015Nobel Prize for neutrino oscillations
- 2026IceCube Upgrade strings deployed
References
- Y. Fukuda et al. (Super-Kamiokande), “Evidence for oscillation of atmospheric neutrinos”, Phys. Rev. Lett. 81, 1562 (1998), arXiv:hep-ex/9807003, doi:10.1103/PhysRevLett.81.1562
- M. Honda, M. Sajjad Athar, T. Kajita, K. Kasahara, S. Midorikawa, “Atmospheric neutrino flux calculation using the NRLMSISE-00 atmospheric model”, Phys. Rev. D 92, 023004 (2015), arXiv:1502.03916, doi:10.1103/PhysRevD.92.023004
- Z. Li et al. (Super-Kamiokande), “Measurement of the tau neutrino cross section in atmospheric neutrino oscillations with Super-Kamiokande”, Phys. Rev. D 98, 052006 (2018), doi:10.1103/PhysRevD.98.052006
- R. Abbasi et al. (IceCube), “Measurement of atmospheric neutrino mixing with improved IceCube DeepCore calibration and data processing”, Phys. Rev. D 108, 012014 (2023), arXiv:2304.12236, doi:10.1103/PhysRevD.108.012014
- S. Aiello et al. (KM3NeT), “Measurement of neutrino oscillation parameters with the first six detection units of KM3NeT/ORCA”, arXiv:2408.07015 (2024), arXiv:2408.07015
- I. Esteban, M. C. Gonzalez-Garcia, M. Maltoni, T. Schwetz, A. Zhou, “NuFit-6.0: updated global analysis of three-flavor neutrino oscillations”, JHEP 12, 216 (2024), arXiv:2410.05380, doi:10.1007/JHEP12(2024)216
Last updated: · Editorial standard
Cite this page
Text
Neutrino Atlas (2026). “Atmospheric neutrinos”. Retrieved 22 November 2026, from https://neutrino-atlas.com/learn/atmospheric-neutrinos/BibTeX
@misc{neutrinoatlas2026earnatmosphericneutrinos,
title = {Atmospheric neutrinos},
author = {{Neutrino Atlas}},
year = {2026},
url = {https://neutrino-atlas.com/learn/atmospheric-neutrinos/},
note = {Accessed 2026-11-22},
language = {english}
}